[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M77XQB8K43FRRANXZP879Z70":3,"skill-same:01M77XQB8K43FRRANXZP879Z70":176},{"skillNo":4,"name":5,"nameZh":6,"description":7,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":16,"categories":17,"keywords":18,"publishedAt":19,"linkOnly":20,"markdownContent":21,"markdownBody":22,"frontMatter":23,"refName":24,"commitSha":25,"skillDirectory":26,"forksCount":27,"repositoryHtmlUrl":28,"repositoryHomepageUrl":6,"ownerHtmlUrl":29,"ownerBlogUrl":6,"ownerDescription":6,"updatedAt":30,"install":31,"resources":44},"01M77XQB8K43FRRANXZP879Z70","vs-shape-it",null,"Use when the user says shape it, brainstorm, explore this idea, grill this, or wants to turn an idea into a buildable design.","vltansky\u002Fvs","vltansky","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F5851280?v=4","01M77XJAE230PB81N52XJHT5A1","Vlad Tansky","01M77XJAET1CMBAZ2TWTT4ZQT8",0,"MIT","https:\u002F\u002Fgithub.com\u002Fvltansky\u002Fvs\u002Fblob\u002Fmaster\u002Fskills\u002Fvs-shape-it\u002FSKILL.md",[],[],"2026-11-15 14:19:17",false,"---\nname: vs-shape-it\ndescription: \"Use when the user says shape it, brainstorm, explore this idea, grill this, or wants to turn an idea into a buildable design.\"\n---\n\n# Shape It\n\nTurn loose ideas and rough plans into approved, buildable designs. Keep the\nconversation short and decision-focused; the human makes strategic calls.\n\n\u003CHARD-GATE>\nShape-it itself never writes code. Until the user approves a READY or READY_WITH_RISKS close,\ndo NOT write code, scaffold projects, create GitHub issues, or start\nimplementation workers. After that approval, shape-it still does not\nimplement: it hands the approved spec to `\u002Fvs-build-it`, or stops or returns to\nits caller, per the Approval table below. Planning-only research or critique subagents are\nallowed for complex work. Output is questions, evidence, design, stress-test,\na decision record when one is warranted, a goal-ready execution blueprint,\nand a closing `\u002Fvs-eli5` HTMDX of the spec. Writing an ADR is a decision record, not implementation, and is allowed. The eli5 is a short review of the spec so the user can confirm it, not a replacement for the spec and not implementation.\n\u003C\u002FHARD-GATE>\n\n## Route the input\n\nInfer the route; do not ask the user to choose a mode.\n\n- **Explore:** vague idea, open question, or no chosen approach. This is the\n  default; follow the workflow below.\n- **Guided Explore:** the user says `interview me`, `ask me questions`,\n  `question me`, or `grill me` and wants to work through their own idea. Follow\n  Explore, but use the adaptive interview cadence in the opening interaction.\n- **Challenge:** formed plan, spec, or RFC, or an explicit request to grill,\n  challenge, or stress-test the proposal itself. Delegate to `\u002Fvs-pushback`; do\n  not duplicate its scoring and verdict workflow here. `Grill this plan` is\n  Challenge; `grill me — question me` is Guided Explore.\n\nIf the initial route was wrong, pivot immediately.\n\nIf the shaped outcome is a vs skill eval contract, compose `\u002Fvs-eval` after\nthe spec is accepted — not during Explore.\n\nAll routes end in pushback. Explore and Guided Explore run it as the last step\nof independent shaping, then the close-time eli5. Challenge hands the whole\nsession to pushback in interactive mode; pushback composes the close-time\n`\u002Fvs-eli5`. Shape-it does not compose a second one.\n\nFor Explore, use the shared\n[`context-docs.md`](..\u002Fvs-internal-shared\u002Freferences\u002Fcontext-docs.md) protocol\nfor project language. Shape-it owns the question rounds, design, ADR outcome,\nand any resolved glossary updates; do not open a separate context interview.\n\nWhen Codex goal tools are available, follow\n[Codex Goal Integration](..\u002Fvs-internal-shared\u002Freferences\u002Fcodex-goal.md).\nShape-it owns only the planning goal, never the later implementation goal.\nFor work that may benefit from delegation, load and follow\n[Subagent Orchestration](..\u002Fvs-internal-shared\u002Freferences\u002Fsubagents.md).\n\n## Long-horizon shaping\n\nShape-it still owns work whose strategic decisions cannot be resolved reliably\nin one session. Treat this as long-horizon shaping, not a separate workflow.\n\nUse the existing design spec as the durable source of truth. Create it during\nindependent shaping when necessary and mark it `Status: SHAPING`. Keep these\nsections current alongside the normal design:\n\n- **Open decisions** — unresolved strategic questions and their dependencies\n- **Evidence** — paths and URLs that constrain those decisions\n- **Next decision** — the single question the next shaping session should resolve\n\nAt a clean boundary, apply\n[Phase Boundaries](..\u002Fvs-internal-shared\u002Freferences\u002Fphase-boundaries.md). Resume\nby loading the spec first, then only the evidence needed for its Next decision;\ndo not reload the whole prior conversation. Keep open decisions out of\nimplementation issues and do not start implementation until the Goal Contract\nis approved. Once the open decisions are resolved, continue with the normal\nspec approval and handoff.\n\n## Explore workflow\n\nThe default cadence is **interaction -> independent -> interaction**. Align\nwhile the user is present, do the substantive shaping without making them\nbabysit the process, then bring back a complete recommendation for decision.\n\n### 1. Opening interaction\n\nAssume the user is available and engaged at the start. Read the sources they\nnamed and inspect only enough nearby context to avoid questions the repository\ncan answer. Do not begin broad research or produce a near-final design before\nalignment.\n\nBefore asking the first question, resolve and read the applicable\n`CONTEXT-MAP.md`\u002F`CONTEXT.md` using the shared context-docs protocol, then use\nits canonical language. Ask about unresolved strategic choices only; do not\nask the user to restate a term the glossary or nearby code already settles.\n\nDiscover the active workspace, repository, and existing artifacts from the host\nand current directory. Ask for an evidence location only when it cannot be\ndiscovered or accessed and choosing the wrong one would materially change the\ndesign; ordinary navigation is not a strategic question.\n\nOpening is **Alignment**. When this phase starts, show the visual progress\ncheckpoint at 25%, with `none yet` for subskills and `Goal Contract + Execution\nStrategy drafting` for Output, per\n[`communication.md`](..\u002Fvs-internal-shared\u002Freferences\u002Fcommunication.md).\n\nEvery Explore session crosses an explicit alignment checkpoint before\nindependent shaping. In the default route, ask up to 3 high-leverage strategic\nquestions in one round. Even when the direction sounds settled, ask one\nquestion about the least-supported expensive-to-reverse assumption instead of\nsilently treating the agent's interpretation as alignment.\n\nSkip the question only when the user explicitly asks for no questions, or the\ncurrent conversation already contains explicit answers for the outcome,\nboundary, and proof of success. In that case, reflect those settled decisions\nin one short alignment statement before independent shaping. Repository facts\nand ordinary navigation never count as user alignment.\n\nWhen choosing questions, use only the lenses that remain unresolved:\n\n- **Problem and outcome** — what should change, for whom, and why it matters\n- **System boundary and ownership** — where the responsibility belongs\n- **Kill criterion** — what evidence would make this not worth building\n\nDo not ask all three by rote or spend a slot on a lens the evidence already\nsettles.\n\n#### Guided Explore interview\n\nUse this cadence only for Guided Explore. It is deliberately more interactive\nthan the default batched checkpoint:\n\n1. Start with a brief assessment of what is known, what remains undecided, and\n   the likely decision areas. Show the visual checkpoint from\n   [`communication.md`](..\u002Fvs-internal-shared\u002Freferences\u002Fcommunication.md).\n   Use this shape at every user-visible phase change:\n\n   ```text\n   ─── \u003CTOPIC> ───\n\n   Progress: \u003C25|50|75|100>% \u003C█░░░|██░░|███░|████>\n   \u003C✓|→|○> Alignment: \u003Cconcrete state>\n   \u003C✓|→|○> Shaping: \u003Cconcrete state>\n   \u003C✓|→|○> Your input needed: \u003Cthe decision, or \"none\">\n   \u003C✓|→|○> Handoff: \u003Cconcrete state>\n   Subskills completed: \u003C\u002Fvs-* → concrete contribution; ...> or \u003Cnone yet>\n   Output: Goal Contract + Execution Strategy \u003Cdrafting|ready|blocked>\n   ```\n\n   The percentage is the current phase ordinal: 25% Alignment, 50% Shaping,\n   75% Your input needed, and 100% Handoff. It is not confidence or a work\n   estimate. Do not name session internals (compact, clear, or the\n   session-action handoff). Do not show a time estimate, target question count,\n   or an empty progress bar.\n2. Ask up to 3 independent consequential strategic questions per round. Batch\n   only questions whose answers do not depend on each other. Lead each with a\n   Why tldr (one line: what the user loses if they pick wrong), then the\n   recommended answer. Always include Drill (`eli5`) as the last option.\n   Compose `\u002Fvs-eli5` on that question's tradeoff only if they pick it.\n3. Branch after each round: choose the next round from its answers, skip\n   decisions those answers settle, and inspect the code or named evidence\n   yourself when a fact can answer the next question.\n4. Accept `done`, `skip`, `back`, `?`, and `eli5`. `skip` accepts the stated\n   reversible default for the current question; `?` defers it as an open\n   decision; `eli5` drills this question's tradeoff with `\u002Fvs-eli5` and waits\n   for A\u002FB after; `back` revisits the previous answer and updates dependent\n   decisions. If the user says `done`, preserve remaining uncertainty as\n   explicit open decisions with recommendations.\n5. Stop interviewing once outcome, boundary, and one expensive-to-reverse\n   call are clear. Do not keep grilling until all expensive choices are clear.\n   Everything else is a stated reversible default in the spec. Keep Why +\n   Drill as the last option. Reflect the agreed mental model in no more than three bullets,\n   then state the aligned model and enter the uninterrupted independent beat\n   without another question. Do not start another question round after that\n   expensive-to-reverse call.\n6. Update progress only when a decision changes state. During alignment, show\n   resolved and open decisions; do not turn progress into a fixed questionnaire.\n\nDo not turn Guided Explore into a fixed questionnaire. Progress means resolved\ndecisions, not a target question count. The user supplies priorities and\ntrade-offs; the agent supplies facts, code reading, and recommendations.\n\nFor either cadence, render questions through the host's structured question\ntool when available\n  (`request_user_input` in Codex, `AskUserQuestion` in Claude Code); see\n  [internal-shared](..\u002Fvs-internal-shared\u002FSKILL.md) Structured questions. Fall\n  back to the Markdown format below otherwise. In Codex, when\n  `request_user_input` is listed, call it rather than rendering the fallback.\n\nFor the default batched checkpoint:\n\n- Batch only independent questions; every question in the round must be\n  answerable without the others' answers. If a question depends on an earlier\n  answer, state the dependency inline or infer a reversible default later.\n- Keep the question turn under about 60 words per question.\n- State the decision in everyday language before presenting options. Make clear\n  what changes for the user; do not make them decode architecture or workflow\n  labels to answer.\n- Every question has a Why tldr: one line, user-world stake if they pick\n  wrong. It is not a call to `\u002Fvs-tldr`.\n- Every question ends with the same Drill option: compose `\u002Fvs-eli5` on\n  that question's tradeoff and open the html. Do not open it unless they\n  pick Drill. In the text fallback it is always `D. Drill — \u002Fvs-eli5 this\n  tradeoff`. In the structured tool it is the last option, same label.\n  `1D` or `eli5` drills that question, then wait for A\u002FB\u002FC.\n- Recommend a path for every choice; put it first and label it the default. In\n  the text fallback, make it option `A`.\n- Accept batched replies like `1A, 2B`; a bare `A` or `yes` accepts every\n  recommendation.\n- Infer tactical details and state safe defaults briefly.\n- Spend one of the three questions on external research only when the design\n  could proceed without it and the lookup would cost real time. When the need is\n  clear, run it in the independent beat instead of asking.\n\nGuided Explore calls the structured tool once per round with one to three\nindependent questions, then waits for all answers. It uses the batched-reply\nshortcut and multi-question fallback below.\n\n```markdown\n## Decisions needed\n\n### 1. \u003Cshort decision title>\n\n**Why:** \u003Cone-line tldr of the user-world stake if they pick wrong>\n\n**Recommended: A — \u003Cchoice>** — \u003Cone-clause rationale>\n\n- A. \u003Cchoice> — \u003Cconsequence>\n- B. \u003Cchoice> — \u003Cconsequence>\n- C. \u003Cchoice> — \u003Cconsequence>\n- D. Drill — `\u002Fvs-eli5` this tradeoff\n\n### 2. \u003Cshort decision title>\n\n...\n\nReply `A` to accept every recommendation, `1D` to drill question 1, or\nspecify changes such as `1B, 2A`.\n```\n\n### 2. Independent shaping\n\nAfter the opening reply, assume the user may be away until the closing\ninteraction. Research, design, challenge, and synthesize autonomously. Do not\ndrip follow-up questions or pause at section-by-section gates. Resolve tactical\nuncertainty with explicit, reversible defaults. Preserve genuinely strategic\nuncertainty as an open decision with a recommendation and consequences, then\ncontinue every safe part of the design.\n\nAutonomous is not silent. Per\n[`..\u002Fvs-internal-shared\u002Freferences\u002Fcommunication.md`](..\u002Fvs-internal-shared\u002Freferences\u002Fcommunication.md),\nthis beat is **Shaping**. When the phase starts, emit the visual progress\ncheckpoint at 50%. Update `Subskills completed` with only finished `\u002Fvs-*`\nskills, each paired with the result Shape It kept, and keep `Output: Goal\nContract + Execution Strategy drafting`. Score\nthe checkpoint with `skills\u002Fvs-shape-it\u002Fscripts\u002Freject-step-map.mjs` (identity\n`ee2e6aed87fd652c945d67a1e7a580bb65ffd23a01a12d7b2b79607ce9a4cdea`;\nexit 1 is a fail). Exclusive cases live under `test\u002Ffixtures\u002Fstep-map`.\n\n```text\n─── STASH CANVAS PILOT ───\n\nProgress: 50% ██░░\n✓ Alignment: pilot goal aligned\n→ Shaping: scope and evidence in progress\n○ Your input needed: none\n○ Handoff: Goal Contract + Execution Strategy\nSubskills completed: none yet\nOutput: Goal Contract + Execution Strategy drafting\n```\n\nDo not turn these into questions. They report state; they do not request input.\nDo not name session internals in the checkpoint. The topic and states above are\nexamples; replace them with the current work.\n\n#### Gather evidence\n\nRead relevant repository docs, nearby code, screenshots, and prior artifacts\nneeded to test the aligned direction. Check\n`~\u002F.vs\u002F$PROJECT_ID\u002F{pushback,specs,context,rfcs}\u002F` when relevant, resolving\n`$PROJECT_ID` per [internal-shared](..\u002Fvs-internal-shared\u002FSKILL.md).\nWhen the topic has project-specific language, also read the applicable\n`CONTEXT.md` or `CONTEXT-MAP.md` using\n[`..\u002Fvs-internal-shared\u002Freferences\u002Fcontext-docs.md`](..\u002Fvs-internal-shared\u002Freferences\u002Fcontext-docs.md).\nIf a term resolves during this session, update the selected glossary inline\naccording to the shared context-docs protocol; do not create an empty context\nfile merely because context support was loaded.\n\nStart with sources the user named and the nearest relevant implementation. Do\nnot inventory the repo or launch broad research until a specific design question\nrequires it. Stop when more reading is unlikely to change the next decision.\n\nFor multi-domain or high-risk work, delegate only independent planning lanes,\nsuch as one bounded evidence inventory and one fresh adversarial critique. The\nparent keeps user alignment, decisions, synthesis, and goal ownership. Do not\ndelegate the interview or ask several agents to design the same whole solution.\n\n#### Architecture evidence\n\nBefore Design, load and follow\n[`..\u002Fvs-architect\u002FSKILL.md`](..\u002Fvs-architect\u002FSKILL.md) in composed mode when all\nof these are true:\n\n- an existing implementation is in scope;\n- the design changes module responsibilities, interfaces, seams, coupling, or\n  consolidation;\n- the recommendation depends on understanding current callers and tests;\n- no current architect result already covers the scoped area.\n\nScope architect to the affected flow or package. It returns evidence and vetted\ncandidates to shape-it; it does not open its standalone candidate-selection gate.\nUse the top candidate as evidence, preserve credible alternatives, and surface\nonly strategic candidate choice through shape-it's normal interaction.\n\nSkip architect for greenfield work, local implementation choices, and an\nalready approved design or spec. A recorded design is the source of truth until\nthe user requests a revision or implementation evidence invalidates it.\n\n#### Research prior art\n\nRepository evidence answers how this codebase works; it cannot tell you whether\nthe approach is the one the ecosystem already settled. Decide once, early in the\nindependent beat, whether an external research flow runs.\n\nRun one without asking when the answer would change the recommendation and you\nare confident which lookup produces it:\n\n- the design depends on an external system's real API, limits, or semantics\n- the user named an external project, tool, standard, or spec\n- the chosen approach has well-known prior art whose tradeoffs are already known\n- a build-vs-adopt call hinges on what an existing library actually does\n\nPick the narrowest flow and follow its SKILL.md file:\n[`..\u002Fvs-octocode\u002FSKILL.md`](..\u002Fvs-octocode\u002FSKILL.md) for prior\nart, ecosystem patterns, and landscape comparison;\n[`..\u002Fvs-rfc-research\u002FSKILL.md`](..\u002Fvs-rfc-research\u002FSKILL.md) when the decision\nis RFC- or ADR-grade and needs code-cited evidence;\n[`..\u002Fvs-steal\u002FSKILL.md`](..\u002Fvs-steal\u002FSKILL.md) for one named external repo.\nScope it to the specific design question, not the whole topic.\n\nOffer it as an opening-round choice when it would cost real time and the design\ncan proceed without it. Skip it, and say so in one clause, when named sources\nand repository evidence already settle the decision — external reading is not a\ndefault tax on every session.\n\nResearch is evidence, not authority. Cite what changed the design, and record a\nprior-art finding you deliberately rejected as a decision with its rationale.\n\n#### Design\n\nLoad and follow [`..\u002Fvs-ponytail\u002FSKILL.md`](..\u002Fvs-ponytail\u002FSKILL.md) in\ncomposed mode while generating candidates, choosing the first slice, and\nfinalizing scope. Start with the first complete rung; do not present a larger\ncandidate unless evidence shows the smaller one cannot deliver the approved\noutcome. Never reduce evidence, verification, safety, or an explicit\nrequirement.\n\nLead with the recommended approach and why. Keep the core chat design under\nabout 450 words, excluding a justified orchestration appendix, and include:\n\n- the smallest valuable, deployable end-to-end slice\n- scope and explicit non-goals\n- terminology and system boundaries\n- data\u002Fcontrol flow and important interfaces\n- failure modes, degradation\u002Frecovery, and ownership when the design crosses a\n  runtime or operational boundary\n- 1-2 alternatives with concrete tradeoffs\n- risks, success criteria, and verification\n\nFor a meaningful solution-size decision, include a compact `### Ponytail cut`\nwith the smallest complete slice, machinery avoided, explicit deferrals, and\nthe evidence that makes the slice complete. Omit it when no machinery or scope\nchoice was made; do not manufacture a section for trivial wording work.\n\nDo not make the first delivery absorb every useful capability discovered during\nshaping. Separate what is required for the smallest complete outcome from later\nreliability, scale, analytics, automation, and platform work. A larger first\nslice needs evidence that the smaller slice cannot produce the approved outcome\nor would create unacceptable risk.\n\nThe slice must advance the approved outcome beyond the confirmed baseline.\nRe-auditing or hardening behavior that already works is not the first delivery\nunless evidence shows that behavior is unreliable and correcting it is part of\nthe user's goal. When new value depends on user access or external ownership,\nsurface that dependency instead of silently substituting lower-value work.\n\nWhen a cheap prototype would answer a costly design question, recommend\n`\u002Fvs-prototype`. When a performance claim shapes the design, use `\u002Fvs-perf` to\ndefine the metric and evaluator before calling it build-ready.\n\n#### Record the decision\n\nThe record precedes the implementation it governs, per\n[Record repo-level decisions before implementation](..\u002F..\u002Fadr\u002Frecord-repo-level-decisions-before-implementation.md).\nResolve the ADR question once, here, with a stated outcome. Silence is not an\nallowed outcome.\n\nAn ADR is warranted when the design settles a decision that is expensive to\nreverse and that future readers will ask \"why did we do it this way\" about:\ncross-cutting workflow or runtime behavior, a convention contributors and agents\nmust follow, an enforcement or validation policy, or a build-vs-adopt call that\nlocks in a dependency or integration boundary. It is not warranted for feature\nwork, local refactors, or tactical implementation choices that a later session\ncan change freely.\n\nFor an ADR connected to a context decision, also follow the context-document\nstorage rules so an existing in-repo or per-user preference is not overridden.\nThe glossary records vocabulary; the ADR records the durable trade-off.\n\nWhen it is warranted, write `adr\u002F\u003Cslug>.md` now, before the closing interaction:\n\n- Follow the repo's existing ADR convention. Absent one, use `adr\u002F` at the repo\n  root with a slug-only filename — lowercase, dash-separated, present-tense\n  imperative, no numeric prefix.\n- Include Date, Context, Decision, Consequences, and the alternatives actually\n  evaluated with the reason each was rejected. Carry over prior art you rejected\n  during research; that rationale is intact now and expensive to reconstruct\n  later.\n- Leave it uncommitted and add no `Status` field. Approval happens at PR merge.\n- Never edit a merged ADR. When this design changes an existing decision, add a\n  new ADR that names the one it supersedes.\n\nWhen the repo has no ADR surface and the decision qualifies, recommend\n`\u002Fvs-setup-adr` in the closing interaction rather than inventing an ADR\nconvention the repo never adopted.\n\nWhen no repo-level decision was settled, say so in one clause with the reason.\n\nEither way, the Goal Contract's `ADR` line records the result.\n\n#### Finalize the spec\n\nDo not restart the interview. Synthesize the conversation and repository\nevidence into the design:\n\n- Small work: the final chat design is enough.\n- Medium or large work: write\n  `~\u002F.vs\u002F$PROJECT_ID\u002Fspecs\u002FYYYY-MM-DD-\u003Ctopic>-design.md`.\n- Include Problem, Solution, Terminology, Boundaries, Decisions, Testing,\n  Out of Scope, Risks, Success Criteria, Goal Contract, and Execution Strategy.\n- In Testing, prefer the highest existing behavioral seam. Name relevant test\n  prior art; propose a new seam only when existing seams cannot prove success.\n\nThe Goal Contract is the stable handoff from shaping to building:\n\n```markdown\n## Goal Contract\n- Implementation objective: \u003Cone observable outcome for \u002Fvs-build-it>\n- Scope: \u003Csurfaces, workflows, files, or systems>\n- Success criteria: \u003Cobservable completion conditions>\n- Verification: \u003Ctests, runtime evidence, CI, review, or acceptance checks>\n- Evidence plan: \u003Csurface + route or command + state or fixture that will prove\n  the outcome, or OPEN DECISION when no surface exists today>\n- ADR: \u003Cpath(s) to the decision record(s) build-it must honor, or None — reason>\n- Context: \u003Cglossary path(s) updated or read, or None — reason>\n- Constraints and approvals: \u003Chard boundaries and human gates>\n```\n\nThe Evidence plan is what the user is really deciding: how the result will be\njudged. Name the concrete thing build-it will point a browser or a command at.\n\"Tests pass\" is a guardrail, not evidence of the outcome.\n\nFor interaction changes, put a small demo scenario table in that Evidence plan:\nscenario, shared route\u002Ffixture\u002Fviewport, action, expected Before\u002FAfter, and the\nassertion that proves the result. Usually 3–4 scenarios cover the main flow and\nthe relevant failure\u002Fretry, cancellation, or boundary case; use fewer for a\nsmall change. Map them to Success criteria, not a separate approval gate. This\nis a capture plan for build-it\u002Fship-it, not permission to record or implement\nduring shaping. Static changes need matched stills, not invented interactions.\n\nIf nothing today can prove the outcome, that is a strategic open decision — the\noptions are building a surface, accepting a weaker proof, or descoping. Carry it\ninto the closing interaction with a recommendation. Do not defer it to build-it\nas a tactical detail; build-it will then head its handoff `UNPROVEN`.\n\nDo not describe activities such as \"implement the plan\" as the objective. State\nthe achieved product or system outcome. A build agent should be able to create\nor reuse its implementation goal from this section without reinterpreting the\nconversation.\n\n#### Design the execution strategy\n\nOrder work by what can invalidate or block delivery, not by architectural\nlayering:\n\n1. Surface strategic decisions, user-provided access, credentials, approvals,\n   and external ownership needed for proof before broad internal work.\n2. Run the cheapest safe test of the assumption most likely to invalidate or\n   reorder the design.\n3. Deliver the smallest deployable vertical slice across the real integration\n   boundary.\n4. Deepen reliability, scale, and optional capabilities only after that slice\n   is observable.\n\n\"Risk first\" means retire delivery uncertainty early; it does not mean build\nthe largest defensive subsystem first. An inventory or foundation is an early\nworkstream only when its result changes a decision or unblocks the vertical\nslice.\n\nEvery design names its execution class:\n\n- **Direct:** one parent, one session, tightly coupled work. No issues or\n  workers. This is the default for small work.\n- **Orchestrated:** multiple independent domains, more than one session, risky\n  evidence collection, or durable coordination. Include the blueprint below.\n\nFor orchestrated work, make the execution plan runnable rather than saying only\n\"use agents\" or \"parallelize\":\n\n1. Choose the source of truth. Use the spec alone for one durable orchestrator;\n   add GitHub issues when work must be independently claimed, reviewed, or\n   resumed by different people\u002Fsessions. Issues describe ready work, not open\n   product decisions.\n2. Choose concrete host primitives. Prefer subagents for bounded internal work\n   under one orchestrator. Recommend Codex tasks\u002Fthreads for durable user-visible\n   ownership or long-running independent work; recommend Claude Code subagents\n   in Claude Code. Do not invent a primitive the host does not provide.\n3. Define workstreams with stable IDs. For each, name its outcome, owner\n   primitive, reasoning effort, inputs, outputs, dependencies, write scope, and\n   verification. Use low effort for deterministic inventory, medium for scoped\n   implementation, and high or xhigh for ambiguous architecture, integration,\n   security, or independent criticism. Name a specific model only when the host\n   supports it and the choice materially helps.\n4. Arrange waves. Parallelize only work with stable inputs and disjoint write\n   scopes. Put foundations, integration, approval, and final verification behind\n   explicit sequential merge gates. Fresh critics should run after evidence is\n   normalized so they do not inherit worker assumptions.\n5. Include copyable worker briefs for every delegated lane: objective, exact\n   scope, evidence paths, constraints, expected return shape, and stop\n   conditions. The parent owns integration, user communication, goal state, and\n   final verification.\n6. Render a Mermaid flowchart when there are at least three workstreams or two\n   execution waves. Show parallel branches, merge gates, approval gates, and the\n   final verification\u002Fhandoff path.\n\nUse this compact table in the spec:\n\n```markdown\n| ID | Outcome | Primitive | Effort | Inputs | Output | Depends on | Write scope | Verification |\n|---|---|---|---|---|---|---|---|---|\n```\n\nThe blueprint is a plan, not authorization. Do not create the issues, tasks,\nthreads, or implementation workers during shape-it.\n\n#### Stress-test with pushback\n\nEvery explore session ends its independent beat by loading\n[`..\u002Fvs-pushback\u002FSKILL.md`](..\u002Fvs-pushback\u002FSKILL.md) and running its composed\nmode over the finished design, Goal Contract, and execution strategy. There is\nno size threshold: a design small enough to survive the grill costs one short\npass, and the sessions that skip it are the ones that needed it.\n\nRun it last, once the design is whole. Grilling a half-formed design produces\nobjections the remaining shaping would have answered anyway.\n\nComposed mode is non-interactive by contract — it must not open question rounds\ninside the independent beat. It answers its own dimensions from the design and\nthe evidence already gathered, records what the evidence cannot settle as\nunresolved with severity, and returns `Verdict`, `Score`, and Top Pushback.\n\nThen integrate before returning:\n\n- fold supported findings into the design, spec, and Goal Contract rather than\n  appending a critique section the user has to reconcile\n- keep a finding you reject, with the reason it does not hold\n- carry an unresolved high-severity finding, or any finding that exposes a\n  strategic contradiction, into the closing interaction as a decision with a\n  recommendation\n\nShape-it still owns synthesis, user communication, and the design. Pushback\nsupplies the verdict, not the plan.\n\n#### Self-review the design\n\nReview the final integrated design, Goal Contract, any ADR, and execution\nstrategy. Fix issues directly rather than handing editorial cleanup to the user:\n\n- **Placeholder scan** — remove any TBD, TODO, placeholder, or incomplete\n  section\n- **Internal consistency** — align the architecture, Goal Contract, ADR, and\n  execution strategy\n- **Scope check** — keep one buildable outcome or decompose work that cannot\n  share one contract\n- **Ambiguity check** — choose one interpretation or surface a strategic open\n  decision\n\nDo not repeat the review after fixing these issues; continue to the closing\ninteraction.\n\n### 3. Closing interaction\n\nExplore and Guided Explore only. Challenge does not compose `\u002Fvs-eli5` here;\npushback does that close.\n\nCite [`..\u002Fvs-internal-shared\u002Freferences\u002Fexplanation-surfaces.md`](..\u002Fvs-internal-shared\u002Freferences\u002Fexplanation-surfaces.md)\nonly for chat TLDR vs artifact, not its opener. The close-time eli5 is required even when that\ncomplexity test would keep the answer in chat: the user is confirming a spec.\n\nOn Explore and Guided Explore, always compose [`\u002Fvs-eli5`](..\u002Fvs-eli5\u002FSKILL.md) on the spec or plan, including\na `NOT_READY` close — the pictures should show the block. Do not skip the eli5\nfor a small or routine shape. The eli5 is a short picture review of that spec\nso the user can confirm it. The Markdown spec, Goal Contract, and any ADR stay\nthe thing they approve and the machine source of truth; link them from the\neli5. Do not treat the eli5 as a replacement contract.\n\nAfter `\u002Fvs-eli5` returns `Saved:`, always produce the eli5 HTML and the chat\nTLDR. Display: save the eli5 HTML and put its path (or “eli5 saved”) in close item 2 as a link; do not auto-open, and do not run `open` or `xdg-open` unless the user picked Drill mid-session or later asks to open the file.\n\nEntering **Your input needed** is a phase change: emit the visual progress\ncheckpoint at 75%, then the close. Name the exact approval or rework decision,\nlist every completed `\u002Fvs-*` subskill with its concrete contribution (including\nthe `\u002Fvs-eli5` review path and `\u002Fvs-pushback` verdict here), and set `Output` to\n`Goal Contract + Execution Strategy ready` or `blocked`. The checkpoint is not\na fifth close item. After approval, emit the 100% **Handoff** checkpoint, then\nfollow the Approval table. On Build, write\n`Handoff: approved spec → \u002Fvs-build-it \u003Cspec path>`. That checkpoint is a\nbridge, not an end: build-it's first `[1\u002F7]` line continues the same run. Do\nnot re-greet, re-summarize the design, or restart the map.\n\nChat is only this exclusive 4-item close, in this order:\n\n1. The first sentence is the TLDR returned by composed `\u002Fvs-eli5` (two to four short lines): the recommendation in plain English and why it is the best fit. Do not write a second TLDR.\n2. The eli5 file path (or “eli5 saved”).\n3. Verdict-honest Handoff. READY or READY_WITH_RISKS: `Handoff: Goal Contract ready | \u003CN> open decisions` (or the missing field). NOT_READY: `Handoff: Goal Contract blocked — not ready, rework` (or the missing field). Never write `Goal Contract ready` on a NOT_READY close.\n4. One `Your action` line. READY or READY_WITH_RISKS: approval. Required shape, by case:\n   - READY: Reply `approve` to start `\u002Fvs-build-it` on this spec, or request changes.\n   - READY_WITH_RISKS: Reply `approve` to accept \u003Cthe main open risk, one clause> and start `\u002Fvs-build-it`, or request changes.\n   - Spec or design only (the user asked for no implementation): Reply `approve` to accept the spec, or `build` to start `\u002Fvs-build-it`.\n   - NOT_READY: rework only — do not offer approve or `\u002Fvs-build-it`.\n\nDo not paste the Goal Contract, ADR, pushback report, execution blueprint, or\nthe `Execution:` block into chat. Those live in the linked files and the eli5.\nRouting metadata does not replace or suppress the closing design.\n\nPut the complete recommendation and all of the following in the linked files, not in chat:\n\n- the complete recommendation, stress-test changes, and any ADR (path plus\n  one-line decision, or one clause why none was warranted)\n- unresolved strategic decisions, each with one recommended path and how\n  alternatives change the outcome\n- the pushback result as the state of the design, one compact sentence plus\n  surviving high and medium findings (`The design is ready to build, with one\n  open rollout risk (READY_WITH_RISKS).`)\n- the sharpest supported reason the recommendation can fail, with cause and user consequence; a generic risk label is not enough\n- a `NOT_READY` verdict is rework only: it blocks the approval gate. The spec\n  and eli5 lead with the blocking finding. Your action is rework, not\n  approve and not `\u002Fvs-build-it`. Offer interactive `\u002Fvs-pushback` if they\n  want to defend it in rounds\n- the smallest handoff that can execute the approved Goal Contract, default\n  `\u002Fvs-build-it`, plus the execution class below\n\nApproval exists only for READY or READY_WITH_RISKS. Keep one approval request\nin those cases. Ask for approval once, after the whole design, Goal Contract, any ADR, and any\nexecution blueprint are visible in those files. If the\ncomposed pushback verdict is NOT_READY, skip the approval gate: Your action\nis rework. If an unresolved strategic decision remains on a READY or\nREADY_WITH_RISKS close and each option is already shaped, combine it with\nthis gate: `approve` accepts the recommended option.\n\n#### Approval\n\nApproval can start an autonomous build, so the trigger is strict. Check the\nnon-Build rows first: any non-Build row that matches wins over Build. Only\nwhen none matches, use a Build row.\n\n| Situation | Outcome |\n| --- | --- |\n| Direct Explore or Guided Explore run, READY, the user did not ask for spec or design only: the reply is `approve` (any case, punctuation ignored), or a bare acceptance such as `approved`, `yes`, or `lgtm`, with no edit, condition, question, or non-recommended pick | **Build:** emit the 100% Handoff, then load and follow `..\u002Fvs-build-it\u002FSKILL.md` on the approved spec path, with the Goal Contract as its contract |\n| Direct Explore or Guided Explore run, READY_WITH_RISKS, the user did not ask for spec or design only: the same strict acceptance, to an approve line that named the risk | **Build:** the risk is now accepted. Emit the 100% Handoff, then load and follow `..\u002Fvs-build-it\u002FSKILL.md` on the approved spec path |\n| The reply carries an edit, condition, or question, for example `approve but rename X` or `looks good, one question` | **Revise:** treat it as a requested change. Revise the design, then close again. Do not build |\n| The reply picks a non-recommended option on the combined decision gate | **Revise:** fold that pick into the design, then close again. Do not build |\n| Ambiguous reply (you cannot tell acceptance from a change) | **Ask:** ask once which it is. Do not build |\n| The user asked for a spec or design only, not implementation | **Finalize:** `approve` accepts the spec and stops. Do not build. A later `build` reply is a new `\u002Fvs-build-it` request |\n| Composed run: a caller workflow loaded shape-it and is still running, waiting for the spec | **Return:** hand the approved spec path back to that caller. The caller owns the build-it handoff, so build-it never runs twice |\n| NOT_READY close | **Rework:** no approve was offered. Revise, or offer interactive `\u002Fvs-pushback` |\n| Challenge route | **Return:** `\u002Fvs-pushback` owns that close. Shape-it does not start build-it |\n| The host cannot resolve `..\u002Fvs-build-it\u002FSKILL.md` | **Stop:** tell the user to type `\u002Fvs-build-it \u003Cspec path>` and stop |\n\nA caller that handed off and ended does not make the run composed. Examples:\n`\u002Fvs-architect` handing the user to shape-it, or `\u002Fvs-build-it` stopping to\nroute a decision here (its architect step). That run is direct: a Build row\napplies, with a fresh build-it run on the approved spec.\n\nOn Build, if shape-it owns a Codex planning goal, mark it complete before\nloading build-it, so build-it's `get_goal` finds no conflicting active goal\n([Codex Goal Integration](..\u002Fvs-internal-shared\u002Freferences\u002Fcodex-goal.md)).\n\nOn Build, the default session action is Continue. Apply\n[Phase Boundaries](..\u002Fvs-internal-shared\u002Freferences\u002Fphase-boundaries.md) only on\nits observable signals: if they choose Clear or Handoff (the spec is durable\nand the shaping context is now a burden), end with\n`Next: \u002Fvs-build-it \u003Cspec path>` and a one-clause reason. Never stop at a TLDR\nthat names `\u002Fvs-build-it` without that reason.\n\nA Build approval authorizes `\u002Fvs-build-it` only. It does not authorize\n`\u002Fvs-to-issues`, Codex tasks\u002Fthreads, or other workers proposed in the\nexecution class below. Each needs its own explicit OK, and then build-it or\n`\u002Fvs-to-issues` creates them, never shape-it. Build-it's own rules still\ngovern how it runs.\n\nOn READY or READY_WITH_RISKS only, default to the approved spec, then `\u002Fvs-build-it`.\nOn NOT_READY, skip that default. Extra coordination follows\nthe execution class written in the spec, not in chat:\n\n- **Durability:** propose issues when work spans sessions or people, needs a\n  shared dependency graph, or must survive chat context. Distinguish unresolved\n  decision issues from ready-to-build implementation issues; route only the\n  latter to `\u002Fvs-to-issues`.\n- **Parallelism:** propose host-native workers only for independent, bounded\n  lanes. In Codex, prefer subagents under one orchestrator; propose tasks\u002Fthreads\n  for durable user-visible ownership. In Claude Code, propose subagents. If the\n  host has no parallel primitive, recommend sequential execution. Name the\n  concrete primitive; do not stop at generic “sessions,” “lanes,” or “agents.”\n- **Both:** issues remain the source of truth; each worker references one issue.\n```text\nExecution: direct | orchestrated — \u003Cwhy>\nSource of truth: spec | spec + GitHub issues — \u003Cwhy>\nRuntime: parent only | parent + subagents | Codex tasks\u002Fthreads | Claude subagents — \u003Cwhy>\nNext: \u002Fvs-build-it with implementation objective: \u003Cobjective>\n```\n\nWrite that block in the spec. Do not print it in chat.\n\n## Confusion\n\nWhen strategic signals conflict during the opening or closing interaction, name\nthe conflict with recommended options. During independent shaping, record the\nconflict, use a clearly provisional recommendation where safe, and continue\nwork that does not depend on it. Stop early only when no useful safe shaping can\ncontinue. Do not silently choose between materially different outcomes.\n\n## Verification\n\nBefore finishing, check:\n\n- before approval, no implementation, issues, tasks\u002Fthreads, or implementation workers were created\n- the reply was matched to the Approval table, non-Build rows first; spec-only, composed, edited, ambiguous, and Challenge replies never reach build-it\n- named sources and enough nearby evidence were read before asking answerable questions\n- external research either ran with its finding cited, or was skipped with a stated reason\n- pushback ran in composed mode over the finished design and returned a verdict\n  and score, and its supported findings were folded into the design\n- the cadence was opening interaction, uninterrupted independent shaping, then closing interaction\n- READY or READY_WITH_RISKS: the design has one approval gate and a behavioral verification seam; NOT_READY has no approval gate\n- the completed design passed placeholder, consistency, scope, and ambiguity review\n- the first delivery is the smallest complete vertical slice, with later\n  capabilities explicitly deferred\n- the first delivery advances the approved outcome beyond the confirmed baseline\n- user-dependent blockers and plan-invalidating assumptions are tested before\n  broad internal foundations\n- the Goal Contract states an observable outcome, scope, success, and proof\n- the Goal Contract names an Evidence plan, or marks it OPEN DECISION and\n  surfaces it in the closing interaction\n- the ADR question was resolved with a stated outcome: a drafted `adr\u002F\u003Cslug>.md`\n  named in the Goal Contract, or a one-clause reason none was warranted\n- resolved project vocabulary was written inline to the applicable glossary,\n  or no context file was created because no term crystallised\n- `CONTEXT.md` contains vocabulary only, with no implementation detail or spec\n- unresolved strategic ambiguity is explicit\n- direct work has no coordination overhead\n- orchestrated work has bounded workstreams, effort, dependencies, merge gates,\n  worker briefs, and a diagram when the topology is non-trivial\n- the handoff is runnable by build-it without relying on hidden chat context\n\nBefore the final handoff, apply\n[Phase Boundaries](..\u002Fvs-internal-shared\u002Freferences\u002Fphase-boundaries.md). Keep\n`Next` below as the semantic route; report a session action only when required\nby that contract.\n\nWhen the shaped design carries multiple options, risks, or non-trivial\ntopology, present it to the user as an HTMDX proposal page through\n[`..\u002Fvs-show-me\u002FSKILL.md`](..\u002Fvs-show-me\u002FSKILL.md) (the proposal shell) with a\nshort chat TLDR — not a wall of chat text. The machine-facing handoff for\nbuild-it stays Markdown. When this run creates HTMDX, inherit the URL +\nfirst-screen shot handoff from `\u002Fvs-show-me`. Pointer only.\n\n## Output style\n\nApply the [shared output style](..\u002Fvs-internal-shared\u002Freferences\u002Foutput-style.md)\nto every user-facing message.\n\n## Workflow\n\nDirect: on a Build approval (Approval table), continue into **Next**; otherwise emit **Next** only. Composed: return the approved spec to the caller.\n\n**Prev:** idea, rough plan, or question\n**Next:** `\u002Fvs-build-it`\n**Relevant:** `\u002Fvs-improve` | `\u002Fvs-eli5`\n","# Shape It\n\nTurn loose ideas and rough plans into approved, buildable designs. Keep the\nconversation short and decision-focused; the human makes strategic calls.\n\n\u003CHARD-GATE>\nShape-it itself never writes code. Until the user approves a READY or READY_WITH_RISKS close,\ndo NOT write code, scaffold projects, create GitHub issues, or start\nimplementation workers. After that approval, shape-it still does not\nimplement: it hands the approved spec to `\u002Fvs-build-it`, or stops or returns to\nits caller, per the Approval table below. Planning-only research or critique subagents are\nallowed for complex work. Output is questions, evidence, design, stress-test,\na decision record when one is warranted, a goal-ready execution blueprint,\nand a closing `\u002Fvs-eli5` HTMDX of the spec. Writing an ADR is a decision record, not implementation, and is allowed. The eli5 is a short review of the spec so the user can confirm it, not a replacement for the spec and not implementation.\n\u003C\u002FHARD-GATE>\n\n## Route the input\n\nInfer the route; do not ask the user to choose a mode.\n\n- **Explore:** vague idea, open question, or no chosen approach. This is the\n  default; follow the workflow below.\n- **Guided Explore:** the user says `interview me`, `ask me questions`,\n  `question me`, or `grill me` and wants to work through their own idea. Follow\n  Explore, but use the adaptive interview cadence in the opening interaction.\n- **Challenge:** formed plan, spec, or RFC, or an explicit request to grill,\n  challenge, or stress-test the proposal itself. Delegate to `\u002Fvs-pushback`; do\n  not duplicate its scoring and verdict workflow here. `Grill this plan` is\n  Challenge; `grill me — question me` is Guided Explore.\n\nIf the initial route was wrong, pivot immediately.\n\nIf the shaped outcome is a vs skill eval contract, compose `\u002Fvs-eval` after\nthe spec is accepted — not during Explore.\n\nAll routes end in pushback. Explore and Guided Explore run it as the last step\nof independent shaping, then the close-time eli5. Challenge hands the whole\nsession to pushback in interactive mode; pushback composes the close-time\n`\u002Fvs-eli5`. Shape-it does not compose a second one.\n\nFor Explore, use the shared\n[`context-docs.md`](..\u002Fvs-internal-shared\u002Freferences\u002Fcontext-docs.md) protocol\nfor project language. Shape-it owns the question rounds, design, ADR outcome,\nand any resolved glossary updates; do not open a separate context interview.\n\nWhen Codex goal tools are available, follow\n[Codex Goal Integration](..\u002Fvs-internal-shared\u002Freferences\u002Fcodex-goal.md).\nShape-it owns only the planning goal, never the later implementation goal.\nFor work that may benefit from delegation, load and follow\n[Subagent Orchestration](..\u002Fvs-internal-shared\u002Freferences\u002Fsubagents.md).\n\n## Long-horizon shaping\n\nShape-it still owns work whose strategic decisions cannot be resolved reliably\nin one session. Treat this as long-horizon shaping, not a separate workflow.\n\nUse the existing design spec as the durable source of truth. Create it during\nindependent shaping when necessary and mark it `Status: SHAPING`. Keep these\nsections current alongside the normal design:\n\n- **Open decisions** — unresolved strategic questions and their dependencies\n- **Evidence** — paths and URLs that constrain those decisions\n- **Next decision** — the single question the next shaping session should resolve\n\nAt a clean boundary, apply\n[Phase Boundaries](..\u002Fvs-internal-shared\u002Freferences\u002Fphase-boundaries.md). Resume\nby loading the spec first, then only the evidence needed for its Next decision;\ndo not reload the whole prior conversation. Keep open decisions out of\nimplementation issues and do not start implementation until the Goal Contract\nis approved. Once the open decisions are resolved, continue with the normal\nspec approval and handoff.\n\n## Explore workflow\n\nThe default cadence is **interaction -> independent -> interaction**. Align\nwhile the user is present, do the substantive shaping without making them\nbabysit the process, then bring back a complete recommendation for decision.\n\n### 1. Opening interaction\n\nAssume the user is available and engaged at the start. Read the sources they\nnamed and inspect only enough nearby context to avoid questions the repository\ncan answer. Do not begin broad research or produce a near-final design before\nalignment.\n\nBefore asking the first question, resolve and read the applicable\n`CONTEXT-MAP.md`\u002F`CONTEXT.md` using the shared context-docs protocol, then use\nits canonical language. Ask about unresolved strategic choices only; do not\nask the user to restate a term the glossary or nearby code already settles.\n\nDiscover the active workspace, repository, and existing artifacts from the host\nand current directory. Ask for an evidence location only when it cannot be\ndiscovered or accessed and choosing the wrong one would materially change the\ndesign; ordinary navigation is not a strategic question.\n\nOpening is **Alignment**. When this phase starts, show the visual progress\ncheckpoint at 25%, with `none yet` for subskills and `Goal Contract + Execution\nStrategy drafting` for Output, per\n[`communication.md`](..\u002Fvs-internal-shared\u002Freferences\u002Fcommunication.md).\n\nEvery Explore session crosses an explicit alignment checkpoint before\nindependent shaping. In the default route, ask up to 3 high-leverage strategic\nquestions in one round. Even when the direction sounds settled, ask one\nquestion about the least-supported expensive-to-reverse assumption instead of\nsilently treating the agent's interpretation as alignment.\n\nSkip the question only when the user explicitly asks for no questions, or the\ncurrent conversation already contains explicit answers for the outcome,\nboundary, and proof of success. In that case, reflect those settled decisions\nin one short alignment statement before independent shaping. Repository facts\nand ordinary navigation never count as user alignment.\n\nWhen choosing questions, use only the lenses that remain unresolved:\n\n- **Problem and outcome** — what should change, for whom, and why it matters\n- **System boundary and ownership** — where the responsibility belongs\n- **Kill criterion** — what evidence would make this not worth building\n\nDo not ask all three by rote or spend a slot on a lens the evidence already\nsettles.\n\n#### Guided Explore interview\n\nUse this cadence only for Guided Explore. It is deliberately more interactive\nthan the default batched checkpoint:\n\n1. Start with a brief assessment of what is known, what remains undecided, and\n   the likely decision areas. Show the visual checkpoint from\n   [`communication.md`](..\u002Fvs-internal-shared\u002Freferences\u002Fcommunication.md).\n   Use this shape at every user-visible phase change:\n\n   ```text\n   ─── \u003CTOPIC> ───\n\n   Progress: \u003C25|50|75|100>% \u003C█░░░|██░░|███░|████>\n   \u003C✓|→|○> Alignment: \u003Cconcrete state>\n   \u003C✓|→|○> Shaping: \u003Cconcrete state>\n   \u003C✓|→|○> Your input needed: \u003Cthe decision, or \"none\">\n   \u003C✓|→|○> Handoff: \u003Cconcrete state>\n   Subskills completed: \u003C\u002Fvs-* → concrete contribution; ...> or \u003Cnone yet>\n   Output: Goal Contract + Execution Strategy \u003Cdrafting|ready|blocked>\n   ```\n\n   The percentage is the current phase ordinal: 25% Alignment, 50% Shaping,\n   75% Your input needed, and 100% Handoff. It is not confidence or a work\n   estimate. Do not name session internals (compact, clear, or the\n   session-action handoff). Do not show a time estimate, target question count,\n   or an empty progress bar.\n2. Ask up to 3 independent consequential strategic questions per round. Batch\n   only questions whose answers do not depend on each other. Lead each with a\n   Why tldr (one line: what the user loses if they pick wrong), then the\n   recommended answer. Always include Drill (`eli5`) as the last option.\n   Compose `\u002Fvs-eli5` on that question's tradeoff only if they pick it.\n3. Branch after each round: choose the next round from its answers, skip\n   decisions those answers settle, and inspect the code or named evidence\n   yourself when a fact can answer the next question.\n4. Accept `done`, `skip`, `back`, `?`, and `eli5`. `skip` accepts the stated\n   reversible default for the current question; `?` defers it as an open\n   decision; `eli5` drills this question's tradeoff with `\u002Fvs-eli5` and waits\n   for A\u002FB after; `back` revisits the previous answer and updates dependent\n   decisions. If the user says `done`, preserve remaining uncertainty as\n   explicit open decisions with recommendations.\n5. Stop interviewing once outcome, boundary, and one expensive-to-reverse\n   call are clear. Do not keep grilling until all expensive choices are clear.\n   Everything else is a stated reversible default in the spec. Keep Why +\n   Drill as the last option. Reflect the agreed mental model in no more than three bullets,\n   then state the aligned model and enter the uninterrupted independent beat\n   without another question. Do not start another question round after that\n   expensive-to-reverse call.\n6. Update progress only when a decision changes state. During alignment, show\n   resolved and open decisions; do not turn progress into a fixed questionnaire.\n\nDo not turn Guided Explore into a fixed questionnaire. Progress means resolved\ndecisions, not a target question count. The user supplies priorities and\ntrade-offs; the agent supplies facts, code reading, and recommendations.\n\nFor either cadence, render questions through the host's structured question\ntool when available\n  (`request_user_input` in Codex, `AskUserQuestion` in Claude Code); see\n  [internal-shared](..\u002Fvs-internal-shared\u002FSKILL.md) Structured questions. Fall\n  back to the Markdown format below otherwise. In Codex, when\n  `request_user_input` is listed, call it rather than rendering the fallback.\n\nFor the default batched checkpoint:\n\n- Batch only independent questions; every question in the round must be\n  answerable without the others' answers. If a question depends on an earlier\n  answer, state the dependency inline or infer a reversible default later.\n- Keep the question turn under about 60 words per question.\n- State the decision in everyday language before presenting options. Make clear\n  what changes for the user; do not make them decode architecture or workflow\n  labels to answer.\n- Every question has a Why tldr: one line, user-world stake if they pick\n  wrong. It is not a call to `\u002Fvs-tldr`.\n- Every question ends with the same Drill option: compose `\u002Fvs-eli5` on\n  that question's tradeoff and open the html. Do not open it unless they\n  pick Drill. In the text fallback it is always `D. Drill — \u002Fvs-eli5 this\n  tradeoff`. In the structured tool it is the last option, same label.\n  `1D` or `eli5` drills that question, then wait for A\u002FB\u002FC.\n- Recommend a path for every choice; put it first and label it the default. In\n  the text fallback, make it option `A`.\n- Accept batched replies like `1A, 2B`; a bare `A` or `yes` accepts every\n  recommendation.\n- Infer tactical details and state safe defaults briefly.\n- Spend one of the three questions on external research only when the design\n  could proceed without it and the lookup would cost real time. When the need is\n  clear, run it in the independent beat instead of asking.\n\nGuided Explore calls the structured tool once per round with one to three\nindependent questions, then waits for all answers. It uses the batched-reply\nshortcut and multi-question fallback below.\n\n```markdown\n## Decisions needed\n\n### 1. \u003Cshort decision title>\n\n**Why:** \u003Cone-line tldr of the user-world stake if they pick wrong>\n\n**Recommended: A — \u003Cchoice>** — \u003Cone-clause rationale>\n\n- A. \u003Cchoice> — \u003Cconsequence>\n- B. \u003Cchoice> — \u003Cconsequence>\n- C. \u003Cchoice> — \u003Cconsequence>\n- D. Drill — `\u002Fvs-eli5` this tradeoff\n\n### 2. \u003Cshort decision title>\n\n...\n\nReply `A` to accept every recommendation, `1D` to drill question 1, or\nspecify changes such as `1B, 2A`.\n```\n\n### 2. Independent shaping\n\nAfter the opening reply, assume the user may be away until the closing\ninteraction. Research, design, challenge, and synthesize autonomously. Do not\ndrip follow-up questions or pause at section-by-section gates. Resolve tactical\nuncertainty with explicit, reversible defaults. Preserve genuinely strategic\nuncertainty as an open decision with a recommendation and consequences, then\ncontinue every safe part of the design.\n\nAutonomous is not silent. Per\n[`..\u002Fvs-internal-shared\u002Freferences\u002Fcommunication.md`](..\u002Fvs-internal-shared\u002Freferences\u002Fcommunication.md),\nthis beat is **Shaping**. When the phase starts, emit the visual progress\ncheckpoint at 50%. Update `Subskills completed` with only finished `\u002Fvs-*`\nskills, each paired with the result Shape It kept, and keep `Output: Goal\nContract + Execution Strategy drafting`. Score\nthe checkpoint with `skills\u002Fvs-shape-it\u002Fscripts\u002Freject-step-map.mjs` (identity\n`ee2e6aed87fd652c945d67a1e7a580bb65ffd23a01a12d7b2b79607ce9a4cdea`;\nexit 1 is a fail). Exclusive cases live under `test\u002Ffixtures\u002Fstep-map`.\n\n```text\n─── STASH CANVAS PILOT ───\n\nProgress: 50% ██░░\n✓ Alignment: pilot goal aligned\n→ Shaping: scope and evidence in progress\n○ Your input needed: none\n○ Handoff: Goal Contract + Execution Strategy\nSubskills completed: none yet\nOutput: Goal Contract + Execution Strategy drafting\n```\n\nDo not turn these into questions. They report state; they do not request input.\nDo not name session internals in the checkpoint. The topic and states above are\nexamples; replace them with the current work.\n\n#### Gather evidence\n\nRead relevant repository docs, nearby code, screenshots, and prior artifacts\nneeded to test the aligned direction. Check\n`~\u002F.vs\u002F$PROJECT_ID\u002F{pushback,specs,context,rfcs}\u002F` when relevant, resolving\n`$PROJECT_ID` per [internal-shared](..\u002Fvs-internal-shared\u002FSKILL.md).\nWhen the topic has project-specific language, also read the applicable\n`CONTEXT.md` or `CONTEXT-MAP.md` using\n[`..\u002Fvs-internal-shared\u002Freferences\u002Fcontext-docs.md`](..\u002Fvs-internal-shared\u002Freferences\u002Fcontext-docs.md).\nIf a term resolves during this session, update the selected glossary inline\naccording to the shared context-docs protocol; do not create an empty context\nfile merely because context support was loaded.\n\nStart with sources the user named and the nearest relevant implementation. Do\nnot inventory the repo or launch broad research until a specific design question\nrequires it. Stop when more reading is unlikely to change the next decision.\n\nFor multi-domain or high-risk work, delegate only independent planning lanes,\nsuch as one bounded evidence inventory and one fresh adversarial critique. The\nparent keeps user alignment, decisions, synthesis, and goal ownership. Do not\ndelegate the interview or ask several agents to design the same whole solution.\n\n#### Architecture evidence\n\nBefore Design, load and follow\n[`..\u002Fvs-architect\u002FSKILL.md`](..\u002Fvs-architect\u002FSKILL.md) in composed mode when all\nof these are true:\n\n- an existing implementation is in scope;\n- the design changes module responsibilities, interfaces, seams, coupling, or\n  consolidation;\n- the recommendation depends on understanding current callers and tests;\n- no current architect result already covers the scoped area.\n\nScope architect to the affected flow or package. It returns evidence and vetted\ncandidates to shape-it; it does not open its standalone candidate-selection gate.\nUse the top candidate as evidence, preserve credible alternatives, and surface\nonly strategic candidate choice through shape-it's normal interaction.\n\nSkip architect for greenfield work, local implementation choices, and an\nalready approved design or spec. A recorded design is the source of truth until\nthe user requests a revision or implementation evidence invalidates it.\n\n#### Research prior art\n\nRepository evidence answers how this codebase works; it cannot tell you whether\nthe approach is the one the ecosystem already settled. Decide once, early in the\nindependent beat, whether an external research flow runs.\n\nRun one without asking when the answer would change the recommendation and you\nare confident which lookup produces it:\n\n- the design depends on an external system's real API, limits, or semantics\n- the user named an external project, tool, standard, or spec\n- the chosen approach has well-known prior art whose tradeoffs are already known\n- a build-vs-adopt call hinges on what an existing library actually does\n\nPick the narrowest flow and follow its SKILL.md file:\n[`..\u002Fvs-octocode\u002FSKILL.md`](..\u002Fvs-octocode\u002FSKILL.md) for prior\nart, ecosystem patterns, and landscape comparison;\n[`..\u002Fvs-rfc-research\u002FSKILL.md`](..\u002Fvs-rfc-research\u002FSKILL.md) when the decision\nis RFC- or ADR-grade and needs code-cited evidence;\n[`..\u002Fvs-steal\u002FSKILL.md`](..\u002Fvs-steal\u002FSKILL.md) for one named external repo.\nScope it to the specific design question, not the whole topic.\n\nOffer it as an opening-round choice when it would cost real time and the design\ncan proceed without it. Skip it, and say so in one clause, when named sources\nand repository evidence already settle the decision — external reading is not a\ndefault tax on every session.\n\nResearch is evidence, not authority. Cite what changed the design, and record a\nprior-art finding you deliberately rejected as a decision with its rationale.\n\n#### Design\n\nLoad and follow [`..\u002Fvs-ponytail\u002FSKILL.md`](..\u002Fvs-ponytail\u002FSKILL.md) in\ncomposed mode while generating candidates, choosing the first slice, and\nfinalizing scope. Start with the first complete rung; do not present a larger\ncandidate unless evidence shows the smaller one cannot deliver the approved\noutcome. Never reduce evidence, verification, safety, or an explicit\nrequirement.\n\nLead with the recommended approach and why. Keep the core chat design under\nabout 450 words, excluding a justified orchestration appendix, and include:\n\n- the smallest valuable, deployable end-to-end slice\n- scope and explicit non-goals\n- terminology and system boundaries\n- data\u002Fcontrol flow and important interfaces\n- failure modes, degradation\u002Frecovery, and ownership when the design crosses a\n  runtime or operational boundary\n- 1-2 alternatives with concrete tradeoffs\n- risks, success criteria, and verification\n\nFor a meaningful solution-size decision, include a compact `### Ponytail cut`\nwith the smallest complete slice, machinery avoided, explicit deferrals, and\nthe evidence that makes the slice complete. Omit it when no machinery or scope\nchoice was made; do not manufacture a section for trivial wording work.\n\nDo not make the first delivery absorb every useful capability discovered during\nshaping. Separate what is required for the smallest complete outcome from later\nreliability, scale, analytics, automation, and platform work. A larger first\nslice needs evidence that the smaller slice cannot produce the approved outcome\nor would create unacceptable risk.\n\nThe slice must advance the approved outcome beyond the confirmed baseline.\nRe-auditing or hardening behavior that already works is not the first delivery\nunless evidence shows that behavior is unreliable and correcting it is part of\nthe user's goal. When new value depends on user access or external ownership,\nsurface that dependency instead of silently substituting lower-value work.\n\nWhen a cheap prototype would answer a costly design question, recommend\n`\u002Fvs-prototype`. When a performance claim shapes the design, use `\u002Fvs-perf` to\ndefine the metric and evaluator before calling it build-ready.\n\n#### Record the decision\n\nThe record precedes the implementation it governs, per\n[Record repo-level decisions before implementation](..\u002F..\u002Fadr\u002Frecord-repo-level-decisions-before-implementation.md).\nResolve the ADR question once, here, with a stated outcome. Silence is not an\nallowed outcome.\n\nAn ADR is warranted when the design settles a decision that is expensive to\nreverse and that future readers will ask \"why did we do it this way\" about:\ncross-cutting workflow or runtime behavior, a convention contributors and agents\nmust follow, an enforcement or validation policy, or a build-vs-adopt call that\nlocks in a dependency or integration boundary. It is not warranted for feature\nwork, local refactors, or tactical implementation choices that a later session\ncan change freely.\n\nFor an ADR connected to a context decision, also follow the context-document\nstorage rules so an existing in-repo or per-user preference is not overridden.\nThe glossary records vocabulary; the ADR records the durable trade-off.\n\nWhen it is warranted, write `adr\u002F\u003Cslug>.md` now, before the closing interaction:\n\n- Follow the repo's existing ADR convention. Absent one, use `adr\u002F` at the repo\n  root with a slug-only filename — lowercase, dash-separated, present-tense\n  imperative, no numeric prefix.\n- Include Date, Context, Decision, Consequences, and the alternatives actually\n  evaluated with the reason each was rejected. Carry over prior art you rejected\n  during research; that rationale is intact now and expensive to reconstruct\n  later.\n- Leave it uncommitted and add no `Status` field. Approval happens at PR merge.\n- Never edit a merged ADR. When this design changes an existing decision, add a\n  new ADR that names the one it supersedes.\n\nWhen the repo has no ADR surface and the decision qualifies, recommend\n`\u002Fvs-setup-adr` in the closing interaction rather than inventing an ADR\nconvention the repo never adopted.\n\nWhen no repo-level decision was settled, say so in one clause with the reason.\n\nEither way, the Goal Contract's `ADR` line records the result.\n\n#### Finalize the spec\n\nDo not restart the interview. Synthesize the conversation and repository\nevidence into the design:\n\n- Small work: the final chat design is enough.\n- Medium or large work: write\n  `~\u002F.vs\u002F$PROJECT_ID\u002Fspecs\u002FYYYY-MM-DD-\u003Ctopic>-design.md`.\n- Include Problem, Solution, Terminology, Boundaries, Decisions, Testing,\n  Out of Scope, Risks, Success Criteria, Goal Contract, and Execution Strategy.\n- In Testing, prefer the highest existing behavioral seam. Name relevant test\n  prior art; propose a new seam only when existing seams cannot prove success.\n\nThe Goal Contract is the stable handoff from shaping to building:\n\n```markdown\n## Goal Contract\n- Implementation objective: \u003Cone observable outcome for \u002Fvs-build-it>\n- Scope: \u003Csurfaces, workflows, files, or systems>\n- Success criteria: \u003Cobservable completion conditions>\n- Verification: \u003Ctests, runtime evidence, CI, review, or acceptance checks>\n- Evidence plan: \u003Csurface + route or command + state or fixture that will prove\n  the outcome, or OPEN DECISION when no surface exists today>\n- ADR: \u003Cpath(s) to the decision record(s) build-it must honor, or None — reason>\n- Context: \u003Cglossary path(s) updated or read, or None — reason>\n- Constraints and approvals: \u003Chard boundaries and human gates>\n```\n\nThe Evidence plan is what the user is really deciding: how the result will be\njudged. Name the concrete thing build-it will point a browser or a command at.\n\"Tests pass\" is a guardrail, not evidence of the outcome.\n\nFor interaction changes, put a small demo scenario table in that Evidence plan:\nscenario, shared route\u002Ffixture\u002Fviewport, action, expected Before\u002FAfter, and the\nassertion that proves the result. Usually 3–4 scenarios cover the main flow and\nthe relevant failure\u002Fretry, cancellation, or boundary case; use fewer for a\nsmall change. Map them to Success criteria, not a separate approval gate. This\nis a capture plan for build-it\u002Fship-it, not permission to record or implement\nduring shaping. Static changes need matched stills, not invented interactions.\n\nIf nothing today can prove the outcome, that is a strategic open decision — the\noptions are building a surface, accepting a weaker proof, or descoping. Carry it\ninto the closing interaction with a recommendation. Do not defer it to build-it\nas a tactical detail; build-it will then head its handoff `UNPROVEN`.\n\nDo not describe activities such as \"implement the plan\" as the objective. State\nthe achieved product or system outcome. A build agent should be able to create\nor reuse its implementation goal from this section without reinterpreting the\nconversation.\n\n#### Design the execution strategy\n\nOrder work by what can invalidate or block delivery, not by architectural\nlayering:\n\n1. Surface strategic decisions, user-provided access, credentials, approvals,\n   and external ownership needed for proof before broad internal work.\n2. Run the cheapest safe test of the assumption most likely to invalidate or\n   reorder the design.\n3. Deliver the smallest deployable vertical slice across the real integration\n   boundary.\n4. Deepen reliability, scale, and optional capabilities only after that slice\n   is observable.\n\n\"Risk first\" means retire delivery uncertainty early; it does not mean build\nthe largest defensive subsystem first. An inventory or foundation is an early\nworkstream only when its result changes a decision or unblocks the vertical\nslice.\n\nEvery design names its execution class:\n\n- **Direct:** one parent, one session, tightly coupled work. No issues or\n  workers. This is the default for small work.\n- **Orchestrated:** multiple independent domains, more than one session, risky\n  evidence collection, or durable coordination. Include the blueprint below.\n\nFor orchestrated work, make the execution plan runnable rather than saying only\n\"use agents\" or \"parallelize\":\n\n1. Choose the source of truth. Use the spec alone for one durable orchestrator;\n   add GitHub issues when work must be independently claimed, reviewed, or\n   resumed by different people\u002Fsessions. Issues describe ready work, not open\n   product decisions.\n2. Choose concrete host primitives. Prefer subagents for bounded internal work\n   under one orchestrator. Recommend Codex tasks\u002Fthreads for durable user-visible\n   ownership or long-running independent work; recommend Claude Code subagents\n   in Claude Code. Do not invent a primitive the host does not provide.\n3. Define workstreams with stable IDs. For each, name its outcome, owner\n   primitive, reasoning effort, inputs, outputs, dependencies, write scope, and\n   verification. Use low effort for deterministic inventory, medium for scoped\n   implementation, and high or xhigh for ambiguous architecture, integration,\n   security, or independent criticism. Name a specific model only when the host\n   supports it and the choice materially helps.\n4. Arrange waves. Parallelize only work with stable inputs and disjoint write\n   scopes. Put foundations, integration, approval, and final verification behind\n   explicit sequential merge gates. Fresh critics should run after evidence is\n   normalized so they do not inherit worker assumptions.\n5. Include copyable worker briefs for every delegated lane: objective, exact\n   scope, evidence paths, constraints, expected return shape, and stop\n   conditions. The parent owns integration, user communication, goal state, and\n   final verification.\n6. Render a Mermaid flowchart when there are at least three workstreams or two\n   execution waves. Show parallel branches, merge gates, approval gates, and the\n   final verification\u002Fhandoff path.\n\nUse this compact table in the spec:\n\n```markdown\n| ID | Outcome | Primitive | Effort | Inputs | Output | Depends on | Write scope | Verification |\n|---|---|---|---|---|---|---|---|---|\n```\n\nThe blueprint is a plan, not authorization. Do not create the issues, tasks,\nthreads, or implementation workers during shape-it.\n\n#### Stress-test with pushback\n\nEvery explore session ends its independent beat by loading\n[`..\u002Fvs-pushback\u002FSKILL.md`](..\u002Fvs-pushback\u002FSKILL.md) and running its composed\nmode over the finished design, Goal Contract, and execution strategy. There is\nno size threshold: a design small enough to survive the grill costs one short\npass, and the sessions that skip it are the ones that needed it.\n\nRun it last, once the design is whole. Grilling a half-formed design produces\nobjections the remaining shaping would have answered anyway.\n\nComposed mode is non-interactive by contract — it must not open question rounds\ninside the independent beat. It answers its own dimensions from the design and\nthe evidence already gathered, records what the evidence cannot settle as\nunresolved with severity, and returns `Verdict`, `Score`, and Top Pushback.\n\nThen integrate before returning:\n\n- fold supported findings into the design, spec, and Goal Contract rather than\n  appending a critique section the user has to reconcile\n- keep a finding you reject, with the reason it does not hold\n- carry an unresolved high-severity finding, or any finding that exposes a\n  strategic contradiction, into the closing interaction as a decision with a\n  recommendation\n\nShape-it still owns synthesis, user communication, and the design. Pushback\nsupplies the verdict, not the plan.\n\n#### Self-review the design\n\nReview the final integrated design, Goal Contract, any ADR, and execution\nstrategy. Fix issues directly rather than handing editorial cleanup to the user:\n\n- **Placeholder scan** — remove any TBD, TODO, placeholder, or incomplete\n  section\n- **Internal consistency** — align the architecture, Goal Contract, ADR, and\n  execution strategy\n- **Scope check** — keep one buildable outcome or decompose work that cannot\n  share one contract\n- **Ambiguity check** — choose one interpretation or surface a strategic open\n  decision\n\nDo not repeat the review after fixing these issues; continue to the closing\ninteraction.\n\n### 3. Closing interaction\n\nExplore and Guided Explore only. Challenge does not compose `\u002Fvs-eli5` here;\npushback does that close.\n\nCite [`..\u002Fvs-internal-shared\u002Freferences\u002Fexplanation-surfaces.md`](..\u002Fvs-internal-shared\u002Freferences\u002Fexplanation-surfaces.md)\nonly for chat TLDR vs artifact, not its opener. The close-time eli5 is required even when that\ncomplexity test would keep the answer in chat: the user is confirming a spec.\n\nOn Explore and Guided Explore, always compose [`\u002Fvs-eli5`](..\u002Fvs-eli5\u002FSKILL.md) on the spec or plan, including\na `NOT_READY` close — the pictures should show the block. Do not skip the eli5\nfor a small or routine shape. The eli5 is a short picture review of that spec\nso the user can confirm it. The Markdown spec, Goal Contract, and any ADR stay\nthe thing they approve and the machine source of truth; link them from the\neli5. Do not treat the eli5 as a replacement contract.\n\nAfter `\u002Fvs-eli5` returns `Saved:`, always produce the eli5 HTML and the chat\nTLDR. Display: save the eli5 HTML and put its path (or “eli5 saved”) in close item 2 as a link; do not auto-open, and do not run `open` or `xdg-open` unless the user picked Drill mid-session or later asks to open the file.\n\nEntering **Your input needed** is a phase change: emit the visual progress\ncheckpoint at 75%, then the close. Name the exact approval or rework decision,\nlist every completed `\u002Fvs-*` subskill with its concrete contribution (including\nthe `\u002Fvs-eli5` review path and `\u002Fvs-pushback` verdict here), and set `Output` to\n`Goal Contract + Execution Strategy ready` or `blocked`. The checkpoint is not\na fifth close item. After approval, emit the 100% **Handoff** checkpoint, then\nfollow the Approval table. On Build, write\n`Handoff: approved spec → \u002Fvs-build-it \u003Cspec path>`. That checkpoint is a\nbridge, not an end: build-it's first `[1\u002F7]` line continues the same run. Do\nnot re-greet, re-summarize the design, or restart the map.\n\nChat is only this exclusive 4-item close, in this order:\n\n1. The first sentence is the TLDR returned by composed `\u002Fvs-eli5` (two to four short lines): the recommendation in plain English and why it is the best fit. Do not write a second TLDR.\n2. The eli5 file path (or “eli5 saved”).\n3. Verdict-honest Handoff. READY or READY_WITH_RISKS: `Handoff: Goal Contract ready | \u003CN> open decisions` (or the missing field). NOT_READY: `Handoff: Goal Contract blocked — not ready, rework` (or the missing field). Never write `Goal Contract ready` on a NOT_READY close.\n4. One `Your action` line. READY or READY_WITH_RISKS: approval. Required shape, by case:\n   - READY: Reply `approve` to start `\u002Fvs-build-it` on this spec, or request changes.\n   - READY_WITH_RISKS: Reply `approve` to accept \u003Cthe main open risk, one clause> and start `\u002Fvs-build-it`, or request changes.\n   - Spec or design only (the user asked for no implementation): Reply `approve` to accept the spec, or `build` to start `\u002Fvs-build-it`.\n   - NOT_READY: rework only — do not offer approve or `\u002Fvs-build-it`.\n\nDo not paste the Goal Contract, ADR, pushback report, execution blueprint, or\nthe `Execution:` block into chat. Those live in the linked files and the eli5.\nRouting metadata does not replace or suppress the closing design.\n\nPut the complete recommendation and all of the following in the linked files, not in chat:\n\n- the complete recommendation, stress-test changes, and any ADR (path plus\n  one-line decision, or one clause why none was warranted)\n- unresolved strategic decisions, each with one recommended path and how\n  alternatives change the outcome\n- the pushback result as the state of the design, one compact sentence plus\n  surviving high and medium findings (`The design is ready to build, with one\n  open rollout risk (READY_WITH_RISKS).`)\n- the sharpest supported reason the recommendation can fail, with cause and user consequence; a generic risk label is not enough\n- a `NOT_READY` verdict is rework only: it blocks the approval gate. The spec\n  and eli5 lead with the blocking finding. Your action is rework, not\n  approve and not `\u002Fvs-build-it`. Offer interactive `\u002Fvs-pushback` if they\n  want to defend it in rounds\n- the smallest handoff that can execute the approved Goal Contract, default\n  `\u002Fvs-build-it`, plus the execution class below\n\nApproval exists only for READY or READY_WITH_RISKS. Keep one approval request\nin those cases. Ask for approval once, after the whole design, Goal Contract, any ADR, and any\nexecution blueprint are visible in those files. If the\ncomposed pushback verdict is NOT_READY, skip the approval gate: Your action\nis rework. If an unresolved strategic decision remains on a READY or\nREADY_WITH_RISKS close and each option is already shaped, combine it with\nthis gate: `approve` accepts the recommended option.\n\n#### Approval\n\nApproval can start an autonomous build, so the trigger is strict. Check the\nnon-Build rows first: any non-Build row that matches wins over Build. Only\nwhen none matches, use a Build row.\n\n| Situation | Outcome |\n| --- | --- |\n| Direct Explore or Guided Explore run, READY, the user did not ask for spec or design only: the reply is `approve` (any case, punctuation ignored), or a bare acceptance such as `approved`, `yes`, or `lgtm`, with no edit, condition, question, or non-recommended pick | **Build:** emit the 100% Handoff, then load and follow `..\u002Fvs-build-it\u002FSKILL.md` on the approved spec path, with the Goal Contract as its contract |\n| Direct Explore or Guided Explore run, READY_WITH_RISKS, the user did not ask for spec or design only: the same strict acceptance, to an approve line that named the risk | **Build:** the risk is now accepted. Emit the 100% Handoff, then load and follow `..\u002Fvs-build-it\u002FSKILL.md` on the approved spec path |\n| The reply carries an edit, condition, or question, for example `approve but rename X` or `looks good, one question` | **Revise:** treat it as a requested change. Revise the design, then close again. Do not build |\n| The reply picks a non-recommended option on the combined decision gate | **Revise:** fold that pick into the design, then close again. Do not build |\n| Ambiguous reply (you cannot tell acceptance from a change) | **Ask:** ask once which it is. Do not build |\n| The user asked for a spec or design only, not implementation | **Finalize:** `approve` accepts the spec and stops. Do not build. A later `build` reply is a new `\u002Fvs-build-it` request |\n| Composed run: a caller workflow loaded shape-it and is still running, waiting for the spec | **Return:** hand the approved spec path back to that caller. The caller owns the build-it handoff, so build-it never runs twice |\n| NOT_READY close | **Rework:** no approve was offered. Revise, or offer interactive `\u002Fvs-pushback` |\n| Challenge route | **Return:** `\u002Fvs-pushback` owns that close. Shape-it does not start build-it |\n| The host cannot resolve `..\u002Fvs-build-it\u002FSKILL.md` | **Stop:** tell the user to type `\u002Fvs-build-it \u003Cspec path>` and stop |\n\nA caller that handed off and ended does not make the run composed. Examples:\n`\u002Fvs-architect` handing the user to shape-it, or `\u002Fvs-build-it` stopping to\nroute a decision here (its architect step). That run is direct: a Build row\napplies, with a fresh build-it run on the approved spec.\n\nOn Build, if shape-it owns a Codex planning goal, mark it complete before\nloading build-it, so build-it's `get_goal` finds no conflicting active goal\n([Codex Goal Integration](..\u002Fvs-internal-shared\u002Freferences\u002Fcodex-goal.md)).\n\nOn Build, the default session action is Continue. Apply\n[Phase Boundaries](..\u002Fvs-internal-shared\u002Freferences\u002Fphase-boundaries.md) only on\nits observable signals: if they choose Clear or Handoff (the spec is durable\nand the shaping context is now a burden), end with\n`Next: \u002Fvs-build-it \u003Cspec path>` and a one-clause reason. Never stop at a TLDR\nthat names `\u002Fvs-build-it` without that reason.\n\nA Build approval authorizes `\u002Fvs-build-it` only. It does not authorize\n`\u002Fvs-to-issues`, Codex tasks\u002Fthreads, or other workers proposed in the\nexecution class below. Each needs its own explicit OK, and then build-it or\n`\u002Fvs-to-issues` creates them, never shape-it. Build-it's own rules still\ngovern how it runs.\n\nOn READY or READY_WITH_RISKS only, default to the approved spec, then `\u002Fvs-build-it`.\nOn NOT_READY, skip that default. Extra coordination follows\nthe execution class written in the spec, not in chat:\n\n- **Durability:** propose issues when work spans sessions or people, needs a\n  shared dependency graph, or must survive chat context. Distinguish unresolved\n  decision issues from ready-to-build implementation issues; route only the\n  latter to `\u002Fvs-to-issues`.\n- **Parallelism:** propose host-native workers only for independent, bounded\n  lanes. In Codex, prefer subagents under one orchestrator; propose tasks\u002Fthreads\n  for durable user-visible ownership. In Claude Code, propose subagents. If the\n  host has no parallel primitive, recommend sequential execution. Name the\n  concrete primitive; do not stop at generic “sessions,” “lanes,” or “agents.”\n- **Both:** issues remain the source of truth; each worker references one issue.\n```text\nExecution: direct | orchestrated — \u003Cwhy>\nSource of truth: spec | spec + GitHub issues — \u003Cwhy>\nRuntime: parent only | parent + subagents | Codex tasks\u002Fthreads | Claude subagents — \u003Cwhy>\nNext: \u002Fvs-build-it with implementation objective: \u003Cobjective>\n```\n\nWrite that block in the spec. Do not print it in chat.\n\n## Confusion\n\nWhen strategic signals conflict during the opening or closing interaction, name\nthe conflict with recommended options. During independent shaping, record the\nconflict, use a clearly provisional recommendation where safe, and continue\nwork that does not depend on it. Stop early only when no useful safe shaping can\ncontinue. Do not silently choose between materially different outcomes.\n\n## Verification\n\nBefore finishing, check:\n\n- before approval, no implementation, issues, tasks\u002Fthreads, or implementation workers were created\n- the reply was matched to the Approval table, non-Build rows first; spec-only, composed, edited, ambiguous, and Challenge replies never reach build-it\n- named sources and enough nearby evidence were read before asking answerable questions\n- external research either ran with its finding cited, or was skipped with a stated reason\n- pushback ran in composed mode over the finished design and returned a verdict\n  and score, and its supported findings were folded into the design\n- the cadence was opening interaction, uninterrupted independent shaping, then closing interaction\n- READY or READY_WITH_RISKS: the design has one approval gate and a behavioral verification seam; NOT_READY has no approval gate\n- the completed design passed placeholder, consistency, scope, and ambiguity review\n- the first delivery is the smallest complete vertical slice, with later\n  capabilities explicitly deferred\n- the first delivery advances the approved outcome beyond the confirmed baseline\n- user-dependent blockers and plan-invalidating assumptions are tested before\n  broad internal foundations\n- the Goal Contract states an observable outcome, scope, success, and proof\n- the Goal Contract names an Evidence plan, or marks it OPEN DECISION and\n  surfaces it in the closing interaction\n- the ADR question was resolved with a stated outcome: a drafted `adr\u002F\u003Cslug>.md`\n  named in the Goal Contract, or a one-clause reason none was warranted\n- resolved project vocabulary was written inline to the applicable glossary,\n  or no context file was created because no term crystallised\n- `CONTEXT.md` contains vocabulary only, with no implementation detail or spec\n- unresolved strategic ambiguity is explicit\n- direct work has no coordination overhead\n- orchestrated work has bounded workstreams, effort, dependencies, merge gates,\n  worker briefs, and a diagram when the topology is non-trivial\n- the handoff is runnable by build-it without relying on hidden chat context\n\nBefore the final handoff, apply\n[Phase Boundaries](..\u002Fvs-internal-shared\u002Freferences\u002Fphase-boundaries.md). Keep\n`Next` below as the semantic route; report a session action only when required\nby that contract.\n\nWhen the shaped design carries multiple options, risks, or non-trivial\ntopology, present it to the user as an HTMDX proposal page through\n[`..\u002Fvs-show-me\u002FSKILL.md`](..\u002Fvs-show-me\u002FSKILL.md) (the proposal shell) with a\nshort chat TLDR — not a wall of chat text. The machine-facing handoff for\nbuild-it stays Markdown. When this run creates HTMDX, inherit the URL +\nfirst-screen shot handoff from `\u002Fvs-show-me`. Pointer only.\n\n## Output style\n\nApply the [shared output style](..\u002Fvs-internal-shared\u002Freferences\u002Foutput-style.md)\nto every user-facing message.\n\n## Workflow\n\nDirect: on a Build approval (Approval table), continue into **Next**; otherwise emit **Next** only. Composed: return the approved spec to the caller.\n\n**Prev:** idea, rough plan, or question\n**Next:** `\u002Fvs-build-it`\n**Relevant:** `\u002Fvs-improve` | `\u002Fvs-eli5`\n",{"name":5,"description":7},"master","72448991571c9e5f0e76cccf189b8522d442e6eb","skills\u002Fvs-shape-it",1,"https:\u002F\u002Fgithub.com\u002Fvltansky\u002Fvs","https:\u002F\u002Fgithub.com\u002Fvltansky","2026-11-06 18:47:54",{"source":8,"skillName":5,"commands":32,"prompt":42,"zipAvailable":43},[33,36,39],{"tool":34,"command":35},"npx","npx skills add vltansky\u002Fvs --skill vs-shape-it",{"tool":37,"command":38},"bunx","bunx skills add vltansky\u002Fvs --skill vs-shape-it",{"tool":40,"command":41},"pnpm","pnpm dlx skills add vltansky\u002Fvs --skill vs-shape-it","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Fvltansky\u002Fvs\u002Fblob\u002Fmaster\u002Fskills\u002Fvs-shape-it\u002FSKILL.md",true,[45,51,57,61,65,69,73,77,81,85,89,93,97,100,104,108,112,116,120,124,128,132,136,140,144,148,152,156,160,164,168,172],{"path":46,"resourceType":47,"resourceTypeDesc":48,"sizeBytes":49,"sourceHtmlUrl":50},"agents\u002Fopenai.yaml","REFERENCE","参考文档",84,"https:\u002F\u002Fgithub.com\u002Fvltansky\u002Fvs\u002Fblob\u002Fmaster\u002Fskills\u002Fvs-shape-it\u002Fagents\u002Fopenai.yaml",{"path":52,"resourceType":53,"resourceTypeDesc":54,"sizeBytes":55,"sourceHtmlUrl":56},"scripts\u002Freject-step-map.mjs","SCRIPT","脚本",3381,"https:\u002F\u002Fgithub.com\u002Fvltansky\u002Fvs\u002Fblob\u002Fmaster\u002Fskills\u002Fvs-shape-it\u002Fscripts\u002Freject-step-map.mjs",{"path":58,"resourceType":53,"resourceTypeDesc":54,"sizeBytes":59,"sourceHtmlUrl":60},"test\u002Fadr-first.static.eval.ts",4388,"https:\u002F\u002Fgithub.com\u002Fvltansky\u002Fvs\u002Fblob\u002Fmaster\u002Fskills\u002Fvs-shape-it\u002Ftest\u002Fadr-first.static.eval.ts",{"path":62,"resourceType":53,"resourceTypeDesc":54,"sizeBytes":63,"sourceHtmlUrl":64},"test\u002Fapprove-handoff.static.eval.ts",11358,"https:\u002F\u002Fgithub.com\u002Fvltansky\u002Fvs\u002Fblob\u002Fmaster\u002Fskills\u002Fvs-shape-it\u002Ftest\u002Fapprove-handoff.static.eval.ts",{"path":66,"resourceType":53,"resourceTypeDesc":54,"sizeBytes":67,"sourceHtmlUrl":68},"test\u002Fcontext-docs.static.eval.ts",1799,"https:\u002F\u002Fgithub.com\u002Fvltansky\u002Fvs\u002Fblob\u002Fmaster\u002Fskills\u002Fvs-shape-it\u002Ftest\u002Fcontext-docs.static.eval.ts",{"path":70,"resourceType":53,"resourceTypeDesc":54,"sizeBytes":71,"sourceHtmlUrl":72},"test\u002Fdesign-quality.static.eval.ts",4768,"https:\u002F\u002Fgithub.com\u002Fvltansky\u002Fvs\u002Fblob\u002Fmaster\u002Fskills\u002Fvs-shape-it\u002Ftest\u002Fdesign-quality.static.eval.ts",{"path":74,"resourceType":47,"resourceTypeDesc":48,"sizeBytes":75,"sourceHtmlUrl":76},"test\u002Ffixtures\u002Fauto-update-project\u002F.claude\u002Fsettings.json",227,"https:\u002F\u002Fgithub.com\u002Fvltansky\u002Fvs\u002Fblob\u002Fmaster\u002Fskills\u002Fvs-shape-it\u002Ftest\u002Ffixtures\u002Fauto-update-project\u002F.claude\u002Fsettings.json",{"path":78,"resourceType":47,"resourceTypeDesc":48,"sizeBytes":79,"sourceHtmlUrl":80},"test\u002Ffixtures\u002Fauto-update-project\u002Fadr\u002Fadopt-adrs-for-repo-level-decisions.md",559,"https:\u002F\u002Fgithub.com\u002Fvltansky\u002Fvs\u002Fblob\u002Fmaster\u002Fskills\u002Fvs-shape-it\u002Ftest\u002Ffixtures\u002Fauto-update-project\u002Fadr\u002Fadopt-adrs-for-repo-level-decisions.md",{"path":82,"resourceType":47,"resourceTypeDesc":48,"sizeBytes":83,"sourceHtmlUrl":84},"test\u002Ffixtures\u002Fauto-update-project\u002Fpackage.json",199,"https:\u002F\u002Fgithub.com\u002Fvltansky\u002Fvs\u002Fblob\u002Fmaster\u002Fskills\u002Fvs-shape-it\u002Ftest\u002Ffixtures\u002Fauto-update-project\u002Fpackage.json",{"path":86,"resourceType":47,"resourceTypeDesc":48,"sizeBytes":87,"sourceHtmlUrl":88},"test\u002Ffixtures\u002Fcaching-project\u002Fdocs\u002Fplan.md",720,"https:\u002F\u002Fgithub.com\u002Fvltansky\u002Fvs\u002Fblob\u002Fmaster\u002Fskills\u002Fvs-shape-it\u002Ftest\u002Ffixtures\u002Fcaching-project\u002Fdocs\u002Fplan.md",{"path":90,"resourceType":47,"resourceTypeDesc":48,"sizeBytes":91,"sourceHtmlUrl":92},"test\u002Ffixtures\u002Fcaching-project\u002Fpackage.json",142,"https:\u002F\u002Fgithub.com\u002Fvltansky\u002Fvs\u002Fblob\u002Fmaster\u002Fskills\u002Fvs-shape-it\u002Ftest\u002Ffixtures\u002Fcaching-project\u002Fpackage.json",{"path":94,"resourceType":53,"resourceTypeDesc":54,"sizeBytes":95,"sourceHtmlUrl":96},"test\u002Ffixtures\u002Fcaching-project\u002Fsrc\u002Fapi.js",582,"https:\u002F\u002Fgithub.com\u002Fvltansky\u002Fvs\u002Fblob\u002Fmaster\u002Fskills\u002Fvs-shape-it\u002Ftest\u002Ffixtures\u002Fcaching-project\u002Fsrc\u002Fapi.js",{"path":98,"resourceType":47,"resourceTypeDesc":48,"sizeBytes":91,"sourceHtmlUrl":99},"test\u002Ffixtures\u002Fperf-project\u002Fpackage.json","https:\u002F\u002Fgithub.com\u002Fvltansky\u002Fvs\u002Fblob\u002Fmaster\u002Fskills\u002Fvs-shape-it\u002Ftest\u002Ffixtures\u002Fperf-project\u002Fpackage.json",{"path":101,"resourceType":47,"resourceTypeDesc":48,"sizeBytes":102,"sourceHtmlUrl":103},"test\u002Ffixtures\u002Fponytail-plan\u002Fpackage.json",80,"https:\u002F\u002Fgithub.com\u002Fvltansky\u002Fvs\u002Fblob\u002Fmaster\u002Fskills\u002Fvs-shape-it\u002Ftest\u002Ffixtures\u002Fponytail-plan\u002Fpackage.json",{"path":105,"resourceType":53,"resourceTypeDesc":54,"sizeBytes":106,"sourceHtmlUrl":107},"test\u002Ffixtures\u002Fponytail-plan\u002Fsrc\u002Fimport-contacts.js",106,"https:\u002F\u002Fgithub.com\u002Fvltansky\u002Fvs\u002Fblob\u002Fmaster\u002Fskills\u002Fvs-shape-it\u002Ftest\u002Ffixtures\u002Fponytail-plan\u002Fsrc\u002Fimport-contacts.js",{"path":109,"resourceType":47,"resourceTypeDesc":48,"sizeBytes":110,"sourceHtmlUrl":111},"test\u002Ffixtures\u002Fpublishing-project\u002Fpackage.json",171,"https:\u002F\u002Fgithub.com\u002Fvltansky\u002Fvs\u002Fblob\u002Fmaster\u002Fskills\u002Fvs-shape-it\u002Ftest\u002Ffixtures\u002Fpublishing-project\u002Fpackage.json",{"path":113,"resourceType":47,"resourceTypeDesc":48,"sizeBytes":114,"sourceHtmlUrl":115},"test\u002Ffixtures\u002Fstep-map\u002Fbad-no-map\u002Fchat.md",89,"https:\u002F\u002Fgithub.com\u002Fvltansky\u002Fvs\u002Fblob\u002Fmaster\u002Fskills\u002Fvs-shape-it\u002Ftest\u002Ffixtures\u002Fstep-map\u002Fbad-no-map\u002Fchat.md",{"path":117,"resourceType":47,"resourceTypeDesc":48,"sizeBytes":118,"sourceHtmlUrl":119},"test\u002Ffixtures\u002Fstep-map\u002Fbag-of-words\u002Fchat.md",109,"https:\u002F\u002Fgithub.com\u002Fvltansky\u002Fvs\u002Fblob\u002Fmaster\u002Fskills\u002Fvs-shape-it\u002Ftest\u002Ffixtures\u002Fstep-map\u002Fbag-of-words\u002Fchat.md",{"path":121,"resourceType":47,"resourceTypeDesc":48,"sizeBytes":122,"sourceHtmlUrl":123},"test\u002Ffixtures\u002Fstep-map\u002Fclean-map\u002Fchat.md",442,"https:\u002F\u002Fgithub.com\u002Fvltansky\u002Fvs\u002Fblob\u002Fmaster\u002Fskills\u002Fvs-shape-it\u002Ftest\u002Ffixtures\u002Fstep-map\u002Fclean-map\u002Fchat.md",{"path":125,"resourceType":47,"resourceTypeDesc":48,"sizeBytes":126,"sourceHtmlUrl":127},"test\u002Ffixtures\u002Fstep-map\u002Fcopy-phrases-skill.md",205,"https:\u002F\u002Fgithub.com\u002Fvltansky\u002Fvs\u002Fblob\u002Fmaster\u002Fskills\u002Fvs-shape-it\u002Ftest\u002Ffixtures\u002Fstep-map\u002Fcopy-phrases-skill.md",{"path":129,"resourceType":47,"resourceTypeDesc":48,"sizeBytes":130,"sourceHtmlUrl":131},"test\u002Ffixtures\u002Fstep-map\u002Ffour-names-no-here\u002Fchat.md",147,"https:\u002F\u002Fgithub.com\u002Fvltansky\u002Fvs\u002Fblob\u002Fmaster\u002Fskills\u002Fvs-shape-it\u002Ftest\u002Ffixtures\u002Fstep-map\u002Ffour-names-no-here\u002Fchat.md",{"path":133,"resourceType":47,"resourceTypeDesc":48,"sizeBytes":134,"sourceHtmlUrl":135},"test\u002Ffixtures\u002Fstep-map\u002Fheadings-four-names\u002Fchat.md",92,"https:\u002F\u002Fgithub.com\u002Fvltansky\u002Fvs\u002Fblob\u002Fmaster\u002Fskills\u002Fvs-shape-it\u002Ftest\u002Ffixtures\u002Fstep-map\u002Fheadings-four-names\u002Fchat.md",{"path":137,"resourceType":47,"resourceTypeDesc":48,"sizeBytes":138,"sourceHtmlUrl":139},"test\u002Ffixtures\u002Fstep-map\u002Fslogan-only-skill.md",135,"https:\u002F\u002Fgithub.com\u002Fvltansky\u002Fvs\u002Fblob\u002Fmaster\u002Fskills\u002Fvs-shape-it\u002Ftest\u002Ffixtures\u002Fstep-map\u002Fslogan-only-skill.md",{"path":141,"resourceType":47,"resourceTypeDesc":48,"sizeBytes":142,"sourceHtmlUrl":143},"test\u002Ffixtures\u002Fsuggestion-project\u002Fpackage.json",159,"https:\u002F\u002Fgithub.com\u002Fvltansky\u002Fvs\u002Fblob\u002Fmaster\u002Fskills\u002Fvs-shape-it\u002Ftest\u002Ffixtures\u002Fsuggestion-project\u002Fpackage.json",{"path":145,"resourceType":53,"resourceTypeDesc":54,"sizeBytes":146,"sourceHtmlUrl":147},"test\u002Finteraction-cadence.static.eval.ts",6628,"https:\u002F\u002Fgithub.com\u002Fvltansky\u002Fvs\u002Fblob\u002Fmaster\u002Fskills\u002Fvs-shape-it\u002Ftest\u002Finteraction-cadence.static.eval.ts",{"path":149,"resourceType":53,"resourceTypeDesc":54,"sizeBytes":150,"sourceHtmlUrl":151},"test\u002Fponytail-composition.eval.ts",2499,"https:\u002F\u002Fgithub.com\u002Fvltansky\u002Fvs\u002Fblob\u002Fmaster\u002Fskills\u002Fvs-shape-it\u002Ftest\u002Fponytail-composition.eval.ts",{"path":153,"resourceType":53,"resourceTypeDesc":54,"sizeBytes":154,"sourceHtmlUrl":155},"test\u002Fresearch-and-pushback.static.eval.ts",4957,"https:\u002F\u002Fgithub.com\u002Fvltansky\u002Fvs\u002Fblob\u002Fmaster\u002Fskills\u002Fvs-shape-it\u002Ftest\u002Fresearch-and-pushback.static.eval.ts",{"path":157,"resourceType":53,"resourceTypeDesc":54,"sizeBytes":158,"sourceHtmlUrl":159},"test\u002Frisk-first.static.eval.ts",1371,"https:\u002F\u002Fgithub.com\u002Fvltansky\u002Fvs\u002Fblob\u002Fmaster\u002Fskills\u002Fvs-shape-it\u002Ftest\u002Frisk-first.static.eval.ts",{"path":161,"resourceType":53,"resourceTypeDesc":54,"sizeBytes":162,"sourceHtmlUrl":163},"test\u002Fshape-it.eval.ts",22223,"https:\u002F\u002Fgithub.com\u002Fvltansky\u002Fvs\u002Fblob\u002Fmaster\u002Fskills\u002Fvs-shape-it\u002Ftest\u002Fshape-it.eval.ts",{"path":165,"resourceType":53,"resourceTypeDesc":54,"sizeBytes":166,"sourceHtmlUrl":167},"test\u002Fshape-it.history.eval.ts",14061,"https:\u002F\u002Fgithub.com\u002Fvltansky\u002Fvs\u002Fblob\u002Fmaster\u002Fskills\u002Fvs-shape-it\u002Ftest\u002Fshape-it.history.eval.ts",{"path":169,"resourceType":53,"resourceTypeDesc":54,"sizeBytes":170,"sourceHtmlUrl":171},"test\u002Fshape-it.real.eval.ts",9501,"https:\u002F\u002Fgithub.com\u002Fvltansky\u002Fvs\u002Fblob\u002Fmaster\u002Fskills\u002Fvs-shape-it\u002Ftest\u002Fshape-it.real.eval.ts",{"path":173,"resourceType":53,"resourceTypeDesc":54,"sizeBytes":174,"sourceHtmlUrl":175},"test\u002Fstep-map.static.eval.ts",4057,"https:\u002F\u002Fgithub.com\u002Fvltansky\u002Fvs\u002Fblob\u002Fmaster\u002Fskills\u002Fvs-shape-it\u002Ftest\u002Fstep-map.static.eval.ts",[177,184,190,195,202,207],{"skillNo":178,"name":179,"nameZh":6,"description":180,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":181,"categories":182,"keywords":183,"publishedAt":19},"01M77XSKWFMPKFXM0311QKEJHH","vs-write","Use when asked to write, rewrite, edit, tighten, simplify, or polish prose for clarity, or to put text in STE, ASD-STE100, or controlled language. Produces plain, direct, accessible copy without losing substance or precision.","https:\u002F\u002Fgithub.com\u002Fvltansky\u002Fvs\u002Fblob\u002Fmaster\u002Fskills\u002Fvs-write\u002FSKILL.md",[],[],{"skillNo":185,"name":186,"nameZh":6,"description":6,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":187,"categories":188,"keywords":189,"publishedAt":19},"01M77XSJNX2V560HSKQKQD67KP","vs-fake-verify","https:\u002F\u002Fgithub.com\u002Fvltansky\u002Fvs\u002Fblob\u002Fmaster\u002Fskills\u002Fvs-verify\u002Ftest\u002Ffixtures\u002Fpath-end-state\u002Fstub-rejector\u002FSKILL.md",[],[],{"skillNo":191,"name":186,"nameZh":6,"description":6,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":192,"categories":193,"keywords":194,"publishedAt":19},"01M77XSHHBEMMYZ6B4PD384NZH","https:\u002F\u002Fgithub.com\u002Fvltansky\u002Fvs\u002Fblob\u002Fmaster\u002Fskills\u002Fvs-verify\u002Ftest\u002Ffixtures\u002Fpath-end-state\u002Fstructure-paste\u002FSKILL.md",[],[],{"skillNo":196,"name":197,"nameZh":6,"description":198,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":199,"categories":200,"keywords":201,"publishedAt":19},"01M77XSE2Y7MW0KP3DTFNEE2E8","vs-verify","Use when asked to verify, prove it works, check completion, or before claiming done. Produces evidence-backed PASS\u002FWARN\u002FFAIL status.","https:\u002F\u002Fgithub.com\u002Fvltansky\u002Fvs\u002Fblob\u002Fmaster\u002Fskills\u002Fvs-verify\u002Ftest\u002Ffixtures\u002Fpath-end-state\u002Fpublished-pair\u002FSKILL.md",[],[],{"skillNo":203,"name":186,"nameZh":6,"description":6,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":204,"categories":205,"keywords":206,"publishedAt":19},"01M77XSAXFQXNE6VCRC7275Z0X","https:\u002F\u002Fgithub.com\u002Fvltansky\u002Fvs\u002Fblob\u002Fmaster\u002Fskills\u002Fvs-verify\u002Ftest\u002Ffixtures\u002Fpath-end-state\u002Fphrase-complete-beside\u002FSKILL.md",[],[],{"skillNo":208,"name":197,"nameZh":6,"description":198,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":209,"categories":210,"keywords":211,"publishedAt":19},"01M77XS6R0E9X1E5KKSS00WYSS","https:\u002F\u002Fgithub.com\u002Fvltansky\u002Fvs\u002Fblob\u002Fmaster\u002Fskills\u002Fvs-verify\u002FSKILL.md",[],[]]