[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M7F910RRMBN04529ZQZZW76Q":3,"skill-same:01M7F910RRMBN04529ZQZZW76Q":46},{"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":25,"commitSha":26,"skillDirectory":27,"forksCount":28,"repositoryHtmlUrl":29,"repositoryHomepageUrl":30,"ownerHtmlUrl":31,"ownerBlogUrl":30,"ownerDescription":6,"updatedAt":32,"install":33,"resources":45},"01M7F910RRMBN04529ZQZZW76Q","plan",null,"Turn a local working spec into a step-by-step implementation plan. Each step names the owning surface and specialist agent and the surface test command; the plan is written into .formula\u002Fwork\u002Fplans\u002F as a gitignored work artifact.","steepy-it\u002Fsteepy-formula","steepy-it","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F299579966?v=4","01M7F8VT4D9JWJZDQFATW4K3YB","Steepy","01M7F8VT4ZR70KR69DV1RR1Q8W",5,"Apache-2.0","https:\u002F\u002Fgithub.com\u002Fsteepy-it\u002Fsteepy-formula\u002Fblob\u002Fmain\u002Fskills\u002Fplan\u002FSKILL.md",[],[],"2026-11-16 10:05:34",false,"---\nname: plan\ndescription: Turn a local working spec into a step-by-step implementation plan. Each step names the owning surface and specialist agent and the surface test command; the plan is written into .formula\u002Fwork\u002Fplans\u002F as a gitignored work artifact.\nuser-invocable: true\n---\n\u003C!-- steepy:manual-handoff:v1:start -->\n## Manual handoff envelope\n\nManual phase boundaries use this literal shape (the envelope is chat text, never a file):\n\n```yaml\nhandoff: steepy-formula\u002Fv1\nnext: \u003Cphase>\nrequired: \u003Crole\u002Fpath map or none>\nonDemand: \u003Crole\u002Fpath map or none>\n```\n\nAll four keys are mandatory and occur once, in that order. Use explicit `none` for an empty map.\nThe phase input role maps are exact and unique: brainstorm has `required: none`, `onDemand: none`;\nplan has required `spec`, `onDemand: none`; Gear-2 direct implement has required `spec`,\n`onDemand: none`; plan-backed implement has required `plan`, on-demand `source-spec`; either\nimplement alternative additionally authorizes exact `progress-ledger` and `task-results` on demand\nwhen resuming;\nregular Gear-3 `review` has required `criteria`, `task-results`, and `branch-diff`, `onDemand: none`;\nfresh `loop-engineer` has required `goal`, `onDemand: none`; resume may additionally authorize exact\n`loop-ledger` on demand; Gear-4 `review` has required `goal` and `loop-ledger`, `onDemand: none`.\nThus Gear-2 brainstorm → implement maps `spec`; brainstorm → plan maps `spec`; plan → implement\nmaps `plan`, `source-spec`; Gear-3 implement → review maps `criteria`, `task-results`, `branch-diff`;\nloop-engineer → review maps `goal`, `loop-ledger`. The two implement role alternatives and the two\nreview role alternatives are mutually exclusive: never merge or infer them. A role occurs only in\nits mapped phase, alternative, and list.\n\nValidate literally before any work-artifact read. Reject duplicate keys, unknown keys, unknown roles,\nmissing keys, a wrong `next`, and unsafe paths: absolute paths, `..` segments, globs, or anything\noutside the repository. Never infer corrections. An explicit malformed or stale binding fails closed\nwithout discovery. Supporting `onDemand` inputs are never preloaded; read one only for a concrete\nnamed missing fact. Stable routing inputs stay outside the envelope and follow bootstrap routing.\nNative invocation spelling comes only from the active harness. No child inherits general\n`.formula\u002Fwork\u002F**` access; pass only its accepted role paths.\n\nEvery new work artifact starts with verdict\u002Fcontract metadata when that phase requires it, followed\nimmediately by this versioned header before the Markdown title:\n\n```yaml\n\u003C!-- steepy-workflow: v1\nphase: \u003Cbrainstorm|plan|implement|review|loop-engineer|goal-contract>\nstatus: \u003CDRAFT|READY|CONSUMED>\nnext: \u003Cphase|none>\nsource: \u003Csafe exact repo-relative path|none>\nconsumed-by: \u003Csafe exact repo-relative path|none>\n-->\n```\n\nHeader fields are mandatory, unique, and literal. `source` and `consumed-by` must agree with the\nadjacent lifecycle transition and authorize no transitive reads. Headerless artifacts are\nignored by recovery and are available only under explicit user audit scope.\n\n### Exact publication pairs\n\nThese are exact resume capabilities, not paths inferred from provenance. The input retains its\nnormal required role; the output role is added to `onDemand`, as are any existing on-demand companions.\nRegular review retains its required criteria and branch diff. Plan resume adds `onDemand.output-plan`;\nregular Gear-3 review resume adds `onDemand.review-report`. Fresh forms above are unchanged.\nGear-4 controller\u002Fledger handling is excluded from this model-based pair repair.\n\n| Route | Input role | Output role | Companion roles | Input phase | Output phase | Output next |\n| --- | --- | --- | --- | --- | --- | --- |\n| plan | spec | output-plan | none | brainstorm | plan | implement |\n| implement-2-direct | spec | task-results | progress-ledger | brainstorm | implement | none |\n| implement-2-plan | plan | task-results | source-spec,progress-ledger | plan | implement | none |\n| implement-3 | plan | task-results | source-spec,progress-ledger | plan | implement | review |\n| review-3 | task-results | review-report | criteria,branch-diff | implement | review | none |\n\n### Publication interruption prefixes\n\nThis table applies only to the exact authorized pair with matching phase, route and provenance.\nValidate the pair before the fresh-only READY input gate; a resume capability is never inferred.\nInput `next` names the consumer phase until consumed; consumption retains it for plan\u002Fimplement\nand sets it to `none` for review. Output `consumed-by` is `none` throughout this producer's run.\nAn unconsumed input must have `consumed-by: none`; a consumed input must name the exact output.\n\n| Input status | Output status | Action |\n| --- | --- | --- |\n| absent | absent | reject |\n| absent | DRAFT | reject |\n| absent | READY | reject |\n| absent | CONSUMED | reject |\n| DRAFT | absent | reject |\n| DRAFT | DRAFT | reject |\n| DRAFT | READY | reject |\n| DRAFT | CONSUMED | reject |\n| READY | absent | fresh-create |\n| READY | DRAFT | resume-work |\n| READY | READY | finish-consumption |\n| READY | CONSUMED | reject |\n| CONSUMED | absent | reject |\n| CONSUMED | DRAFT | reject |\n| CONSUMED | READY | no-op |\n| CONSUMED | CONSUMED | reject |\n\n`fresh-create` permits a new absent output only in a fresh run, never a missing resume target.\n`resume-work` requires exact output capability and leaves the input READY while work\u002Fgates remain.\nFor `finish-consumption`, revalidate output.source against the exact input path, input.consumed-by\nas `none`, and the recorded approval and verification bound to those input\u002Foutput paths and content\n(excluding lifecycle fields). Recheck current branch\u002FHEAD or uncommitted diff identity where applicable.\nOnly then perform the input-consumption write with input.consumed-by equal to the exact output path.\nFor `no-op`, revalidate the recorded proof with the final reciprocal links instead of the\nunconsumed-input condition; perform no writes or redispatch.\nA missing capability, mismatched path\u002Fphase\u002Fnext\u002Fcontent, absent approval\u002Fproof, or impossible state\nmust fail closed, preserving both files. A READY marker alone is not proof. Do not follow links to\nundeclared evidence; request the exact missing capability or user-delimited scope instead.\nNever select a most-recent output or recover by directory\u002Fbody search.\n\nPublication is ordered: record the phase's applicable approval and verification in the output,\npublish output READY by per-file atomic replacement, then consume the input by per-file atomic\nreplacement. This is not a multi-file atomic operation; READY\u002FREADY is its sole verified intermediate\nprefix. Resume repairs no other state and never repeats completed work or release actions.\nIn manual drive you are the enforcement: no parser, manifest, or validator checks this contract.\n\u003C!-- steepy:manual-handoff:v1:end -->\n\n\n# plan\n\nTurn an approved spec into an implementation plan whose every task states the change it makes, not\nthe goal it chases, with review checkpoints.\n\n> **Engine root:** this skill's base directory is `\u003Cengine-root>\u002Fskills\u002Fplan\u002F`; engine\n> scripts live two levels up, at `\u003Cengine-root>\u002Fscripts\u002F`. Resolve them relative to the base\n> directory your harness reports for this skill.\n\n## Checklist\n0. In autopilot, read the supplied phase manifest first; then read the gear and scale ceremony to it (Step 0).\n1. Load the spec and resolve the graph context (`context resolve`) (Step 1).\n2. Map files and tasks.\n3. Bind each task to the hub (surface, specialist agent, test command, complexity).\n4. Write the plan into `.formula\u002Fwork\u002Fplans\u002F`.\n4.5. Self-review the plan inline — free gate, gear 2 and 3 (Step 4.5).\n5. Validate with `steepy-formula graph validate` (deterministic gate).\n6. Human review gate — ask the user to review the written plan before handoff (gear 3).\n7. Hand off to the `implement` skill.\n\n## Procedure\n\n### Step 0 — Read the gear\n\n**Autopilot preflight:** When the conductor's `phasePrompt` supplies a phase manifest, read that\nmanifest before any other task input. Validate its role and scope against the plan phase and the\nsupplied correlation identity. Eagerly read every `required` input before acting. Do not preload `onDemand`;\nconsult it only for a concrete named missing fact and record the read in the plan phase\noutput\u002Freport. Phase-manifest authority applies only when `drive: autopilot`; manual-drive discovery\nremains Step 1's behavior.\n\n**Manual\u002Fno-manifest preflight — before the gear read:** Accept the exact `spec` capability before any verdict or spec-body read. Manual drive has two exclusive manual entry paths:\n\nFor an explicit resume, also require `onDemand.output-plan` in the envelope (or an exact labeled\n`output-plan` alongside the labeled spec). Validate that exact input\u002Foutput pair using the common\npublication table before the fresh-only READY input gate below: READY\u002FDRAFT resumes work,\nREADY\u002FREADY is finish-consumption, and CONSUMED\u002FREADY is no-op after proof revalidation. No other\nstate is accepted. The output header must be `phase: plan`, `next: implement`, with `source` equal\nto the accepted spec and `consumed-by: none`; a consumed spec must name this exact output.\nRead the output body only for the named missing approval\u002Fproof or draft-progress facts, record that\nreason there when resuming work, and do not reopen sibling work. A verified terminal pair goes\nstraight to Step 7 after any sole consumption repair, without writing another plan or repeating gates.\n\n1. **Explicit binding.** If the user supplied a `spec` handoff envelope or an exact spec path,\n   validate the literal binding and path safety before its header read. The path must be an exact\n   safe repository-relative path and the role must be exactly `spec`; then inspect only its bounded\n   `steepy-workflow` header and require `phase: brainstorm`, `status: READY`, and `next: plan`.\n   Exact valid input is accepted. Missing, malformed, stale, wrong-state, or wrong-role input fails\n   closed: report the literal defect and stop without discovery or inferred correction.\n2. **Bounded recovery.** Only when no envelope or path was supplied, enumerate the immediate specs\n   directory (`.formula\u002Fwork\u002Fspecs\u002F`, no recursion), inspect only bounded `steepy-workflow` headers,\n   list candidates whose header is exactly `phase: brainstorm`, `status: READY`, `next: plan` with\n   path and compact header metadata, and require user selection even when there is exactly one candidate.\n   Never open or read a candidate body. Headerless artifacts are ignored by\n   recovery. If no candidate exists, tell the user to run the `brainstorm` skill (invoke it the way\n   the active harness invokes skills) first and stop. After selection, obtain and validate the exact\n   binding before continuing.\n\nDo not read the selected spec's verdict or body until that exact binding is accepted.\nRead the verdict artifact (`verdict` + `gear`) only from the accepted `.formula\u002Fwork\u002F` spec head.\n- **Artifact present:** honor its gear.\n- **Artifact present with `gear: 4`:** this task belongs to the autonomous loop; bounce to\n  the `loop-engineer` skill and stop.\n- **Artifact with `drive: autopilot` (gear 3):** this run is an unattended autopilot phase. Confirm\n  that the contract agrees with the conductor-supplied phase manifest already read first. Apply the\n  Step 6 autopilot branch below and never ask the user anything.\n  An invalid manifest, missing required input, undeclared or unwritable output, or anything else\n  unresolvable is a manifest\u002Foutput failure: append `\u003CISO> — plan — BLOCKED — run-id=\u003Cconductor-supplied> attempt=\u003Cpositive supplied> \u003Creason>` to `.formula\u002Fwork\u002Ftasks\u002F\u003Cspec-basename>\u002Fautopilot-status.md`\n  and exit non-zero.\n- **Artifact absent and this is gear-1\u002F2 work:** recompute the coverage verdict inline (COVERED \u002F\n  CONFLICT \u002F GAP with `file:line` evidence → gear). Declare the recomputed\n  verdict, gear, and evidence, then ask the user for one confirmation before the ceremony starts —\n  the user may force a different gear, and the override is recorded in the verdict artifact. Then\n  proceed.\n- **Artifact absent and full gear-3 ceremony is required for this command:** do not proceed\n  blind — tell the user to enter through the upstream workflow chain first, and stop.\n\nScale this skill's ceremony to the gear (see \"Ceremony by gear\" below). This skill dispatches **no\nsubagent reviewer**: a plan is prose, gated by the inline self-review (Step 4.5) plus a human review\nof the written plan (Step 6), per the \"Review architecture (gear 3)\" governance rule —\nrecorded as a Rule in the canonical graph.\n\n**Autopilot correlation.** Before appending any current-protocol status marker, use the `run-id` and\npositive `attempt` supplied by the conductor's `phasePrompt`; that correlated identity is authoritative.\nDo not derive either value from filenames. Do not invent a correlation identity. A child without supplied run\u002Fattempt identity\nis unresolvable: exit non-zero without appending an uncorrelated new-protocol marker. Only versioned, correlated status is supported; uncorrelated markers never complete a phase.\n\n**Session declaration (declare, never block).** Say whether this conversation started clean for this\nphase or has already run an earlier phase of the chain in the same session. A same-session phase\nalready grew the context, and every turn of this phase now pays to carry it — name that cost, but\nnever refuse or gate on it: this is a declaration about the conversation, not a measurement, and\nsound work proceeds either way. If the session's history cannot be determined with certainty, say so\nand proceed. Under `drive: autopilot` this declaration is a no-op: the conductor\n(`scripts\u002Fautopilot.mjs`) already opens a fresh headless session per phase, so the discipline stated\nhere is already enforced by that mechanism, not by this prose.\n\n### Ceremony by gear\n\n- **Gear 1**: no plan needed — say so and stop.\n- **Gear 2**: an optional light plan is enough; self-review it (Step 4.5); skip only the human\n  question in Step 6, then perform the shared lifecycle transition.\n- **Gear 3**: run the full procedure, including the Step 4.5 self-review and the Step 6 human gate.\n- **Gear 4**: not this skill's entry point — bounce to the `loop-engineer` skill and stop.\n\n### Step 1 — Load the spec and the hub\n\n**Manual drive:** after Step 0 accepts the exact spec capability, continue with that path only.\nAfter stable bootstrap orientation, eagerly read only the accepted spec body, then resolve the\ngraph context. Never open a sibling or linked work artifact. The accepted `source`\u002F`consumed-by`\nfields authorize no transitive read.\n\n**Hub-mode declaration (once per run).** Run `steepy-formula hub-mode` and carry its declared\nmode — `graph`, `migrate`, or `none`, each printed with one `evidence:` line — through the whole\nrun. The command always exits 0: it is a declaration, not a gate, and this prose never\nre-implements its detection — the command is the single authority. The declared mode branches\nthe manual graph reads (autopilot's inventory is the phase manifest, below):\n\n- **`graph` mode:** resolve the hub context from the canonical graph — run\n  `steepy-formula context resolve --path \u003Cp>...` for the spec's exact paths (exit 0 emits\n  `context.json` + `context.md`). The resolved context supplies each owning **surface**, its\n  Rules — the standards-equivalent — its Terms, and each surface's test command from its Check\n  item's `ext_v1_command` extension property (fall back to the verification Rule's statement\n  prose when the Check carries no command). The bound specialist agent for surface `\u003Cname>` is\n  `.claude\u002Fagents\u002F\u003Cname>-agent.md` by naming convention. On exit 10 UNRESOLVED — normal in a freshly graph-first-initialized repo, where no\n  topology bridge links the detected Packages to the curated Components yet — fall back to\n  the deterministic surface targets:\n  `steepy-formula context resolve --target apex:\u003CprojectId>:surface:\u003Cslug(name)>` for each\n  owning surface, every part derived, never guessed — surface names from\n  `.claude\u002Fagents\u002F\u003Cname>-agent.md` (CLAUDE.md's Surfaces section is the fallback listing),\n  the project id from the hub-mode declaration's `hub-mode '\u003CprojectId>'` echo, `slug(name)`\n  lowercase with non-alphanumeric runs collapsed to `-` — and never target the bare project\n  root on its own: it resolves empty. If the resolve fails, diagnose with\n  `steepy-formula graph status` — exit 3,\n  `initialized: no`, or an engine failure means the graph binding is missing or failed: tell the\n  user to run the `init` skill first and stop.\n- **`migrate` mode:** the repo still carries a pre-graph Markdown hub.\n  Stop and tell the user to run the `graph-migrate` skill (invoke it the way your harness invokes skills) before any other workflow.\n- **Mode `none`:** no graph — the repo is not initialized: tell the user to run\n  the `init` skill first and stop.\n\n**Autopilot:** use the validated phase manifest loaded in Step 0 as the authoritative input\ninventory. Do not independently rediscover or preload inputs beyond its `required` list; use the\nrecorded `onDemand` protocol only when a concrete fact is missing.\n\n### Step 2 — Map files and tasks\n\nList the files each task creates\u002Fmodifies (exact paths). A task is mis-cut if its implementer must **discover** what to change: when its requirements can be stated only as a goal (\"make X work,\" \"write the section on Y\") and not as a sketch of a diff to exact paths, the discovery is missing and belongs to `plan` — split the task, or add a discovery task upfront whose deliverable is the missing facts. This is the other half of what a task already needs: not just the exact paths, but the expected change at each path. Decompose into tasks, each ending in an independently testable deliverable.\n\n**Wide refactors — expand → migrate → contract.** A **wide refactor** — one mechanical change\nwhose blast radius breaks callers across the whole codebase — must not be forced into tasks\nthat cannot land green. Sequence it: **expand** (add the new form beside the old — nothing\nbreaks), **migrate** (move call sites in batches sized by blast radius; each batch is an\nordinary sequential task, suite green between batches), **contract** (delete the old form when\nno caller remains). No dependency-graph machinery — batches are ordinary sequential plan\ntasks. \"Wide\" is strict: if the change lands green in one task, it is not wide.\n\n### Step 3 — Bind each task to the hub\n\nThe plan is the complete spec-to-brief bridge. Every task uses this self-contained format:\n\nStart every task with the canonical boundary `## Task \u003Cpositive integer> — \u003Cshort title>`; task\nintegers are unique and increase in execution order. Under it use the exact labels below. The\ncanonical criterion form is `- **Success criteria:** SC1, SC2`; the parser rejects alternate field names such as `Spec criteria`.\n\n- **Requirements and deliverables:** exact behavior to implement and the independently testable result.\n- **Relevant global constraints:** every spec-wide rule that applies to this task, repeated here rather than assumed.\n- **Surface:** `\u003Csurface>` — from the resolved context.\n- **Specialist agent:** the `\u003Csurface>-agent` that owns implementation — bound as `.claude\u002Fagents\u002F\u003Csurface>-agent.md` by naming convention.\n- **Exact paths:** every created or modified repository-relative path; paths must be safe and exact.\n- **Test command:** the surface's exact command — the resolved context Check's `ext_v1_command` (falling back to the verification Rule's statement prose).\n- **Dependencies:** prior task numbers, or `none`, in executable order.\n- **Complexity:** `mechanical | integration | design` — mechanical = 1-2 files with a complete\n  spec; integration = multi-file or judgment; design = architecture, high-risk, or subtle.\n- **Success criteria:** every mapped success-criterion ID from the source spec.\n\nInclude all exact requirements and deliverables needed to materialize the downstream task brief.\nNo requirement may live only in conversation. If copying a large upstream section would make the\ntask unwieldy, reference it by canonical path + heading and state exactly which requirement it\nsupplies. `implement` reads this line (`Complexity`) to pick the model tier and decide whether the task gets its\nown reviewer (Step 3.4).\n\n### Step 4 — Write the local working plan\n\n**Manual drive:** on a fresh run, choose an absent `.formula\u002Fwork\u002Fplans\u002FYYYY-MM-DD-\u003Ctopic>.md` path;\nan existing target requires the exact resume capability before reading or overwriting it. On resume,\nretain the accepted `output-plan` path, never recalculate its date. After any verdict\u002Fcontract metadata and before the title, write:\n\n```yaml\n\u003C!-- steepy-workflow: v1\nphase: plan\nstatus: DRAFT\nnext: implement\nsource: \u003Cexact-spec-path>\nconsumed-by: none\n-->\n```\n\nThen link back to its source spec with a relative work-area link:\n\n`> Source spec: [\u003Ctopic>](..\u002Fspecs\u002F\u003CYYYY-MM-DD-topic>.md)`\n\n**Autopilot:** write the plan exactly to the single path declared by the authoritative phase\nmanifest's `outputs` list. Do not recompute the date or filename. Use that actual path in the\ncorrelated `DONE` marker, and derive the relative source-spec back-link from its location.\n\nCreate `.formula\u002Fwork\u002Fplans\u002F` if it does not exist. Do not create a plans sub-index. Plans are local workflow artifacts, not stable hub documentation.\n\n### Step 4.5 — Self-review the plan (free gate)\n\n**Gear 2 and 3** — skip only at gear 1 (no plan is produced there).\n\nRead the plan you just wrote with fresh eyes and fix any of these inline — no re-review needed, just fix and move on:\n1. **Every mandatory field** — each task contains exact requirements\u002Fdeliverables, relevant global constraints, surface, specialist agent, exact paths, test command, dependencies, complexity, and success-criterion IDs; no placeholder (`TBD`, `TODO`) remains.\n2. **Binding consistency** — every surface is a known surface among the resolved context's Surfaces, every specialist is `.claude\u002Fagents\u002F\u003Csurface>-agent.md` by naming convention, and every exact test command matches the Check's `ext_v1_command` (or its verification Rule's statement fallback).\n3. **Dependency ordering** — dependencies name existing earlier tasks, ordering is executable, and each task ends in an independently testable deliverable.\n4. **Complexity present** — validate allowed complexity: each `Complexity` value is exactly `mechanical | integration | design` and matches the task text (a multi-file task marked `mechanical` is a bug).\n5. **Criteria coverage** — every requirement and success criterion in the spec maps to at least one task; list and close any gap.\n6. **Safe repository-relative paths** — every path is exact, stays inside the repository, and is owned by the task's declared surface.\n7. **Context independence** — a downstream agent can implement the task from the plan\u002Ftask brief without chat history; large references use canonical path + heading and state the supplied requirement.\n8. **Discovery-free tasks** — no task's requirements are stated only as a goal (\"make X work,\" \"write the section on Y\"); each names exact paths and the expected change at each path, or is split \u002F preceded by a discovery task whose deliverable is the missing facts.\n\nAll checks must pass before appending autopilot `DONE` or asking for human approval.\n\n### Step 5 — Validate (deterministic gate)\n\nRun the read-only graph gate:\n\n```bash\nsteepy-formula graph validate\n```\n\nIt MUST exit 0 with `violations: 0` and a clean revision-chain replay. This skill never edits the graph, so any violation is pre-existing drift: report it and point the user to the `check` skill — do not fix it mid-plan.\n\n**Autopilot:** Before appending `DONE`, run the same canonical plan parse that the implement\nphase will use:\n\n```bash\nnode \u003Cengine-root>\u002Fscripts\u002Fautopilot-context.mjs --verify-plan --repo-root . --plan \u003Cauthoritative-manifest-output-path>\n```\n\nIt MUST print `plan OK`. A rejection is a fixable plan error: repair the plan, repeat the Step 4.5\nself-review, and rerun both deterministic gates. Never append `DONE` for a plan this command has\nnot accepted.\n\n### Step 6 — Human review gate and plan lifecycle\n\n**Gear 3 only — human review gate.** A plan is prose. Every task it defines will pass under the per-task and\nwhole-branch code reviewers downstream, but the plan's *shape* — decomposition, spec coverage, task\nordering — is best caught by a human now. Do **not** dispatch a subagent. Tell the user:\n\n> \"Plan written to `\u003Cpath>` — please review it and tell me if anything should change before we implement.\"\n\nWait for the response. On requested changes, edit the plan and re-run the Step 4.5 self-review.\nProceed only once the user approves.\n\n**Gear 2 gate.** Ask no human-review question; the completed Step 4.5 self-review is the applicable\napproval gate.\n\n**Both gears.** Record the applicable approval and verification (Step 4.5, Step 5, and the human\napproval or autopilot authorization) bound to the exact source\u002Foutput paths and content before READY.\nOnly after those gates succeed, first atomically replace the plan to `status: READY` while retaining\n`source: \u003Cexact-spec-path>`; then atomically replace the source spec to `status: CONSUMED` with\n`consumed-by: \u003Cplan-path>`. Before publication, interruption leaves the spec READY and plan DRAFT;\nbetween the two replacements, READY\u002FREADY is the sole repairable prefix. Use only the exact resume\ncapabilities and common proof checks to finish consumption; a finalized pair is an idempotent no-op.\nNever repair a CONSUMED-spec\u002FDRAFT-plan pair or consume an unverified plan.\n\nAutopilot status timestamps must use UTC with three millisecond digits (`YYYY-MM-DDTHH:mm:ss.sssZ`), generated with `new Date().toISOString()`.\n\n**Autopilot:** If `drive: autopilot` is present in the verdict contract (Step 0), the human gate is superseded — this is a non-blocking checkpoint. The plan file is on disk; the human can interrupt the run. Skip the question, do NOT wait, and append `\u003CISO timestamp> — plan — DONE — run-id=\u003Cconductor-supplied> attempt=\u003Cpositive supplied> \u003Cplan path>` to `.formula\u002Fwork\u002Ftasks\u002F\u003Cspec-basename>\u002Fautopilot-status.md` (spec filename without `.md`; the status file already exists). Step 4.5 self-review and Step 5 `graph validate` still run and still gate.\n\n### Step 7 — Hand off\n\nOnly after the applicable Gear-2 or Gear-3 gate and lifecycle transition succeed, tell the user the next step is the `implement` skill, using the\nnative invocation spelling only when the active harness provides it. In manual drive emit:\n\n```yaml\nhandoff: steepy-formula\u002Fv1\nnext: implement\nrequired:\n  plan: \u003Cplan-path>\nonDemand:\n  source-spec: \u003Cspec-path>\n```\n\nRun the `implement` skill in a new session: this phase's context does not serve implementation, and every turn of the implement phase would pay to carry it forward.\n\n## Model Selection\n\nAlways set `model:` explicitly when dispatching a subagent. An omitted model inherits the session model (usually the most capable and most expensive) and silently defeats this policy.\n\nPick the tier by task complexity and risk:\n\n| Signal | Tier |\n|--------|------|\n| Mechanical \u002F transcription (1-2 files, complete spec) | cheap |\n| Integration \u002F multi-file \u002F judgment | standard |\n| Design \u002F architecture \u002F high-risk or subtle change | most-capable |\n\nReviewers floor at **standard** and rise to **most-capable** when the artifact is high-risk or subtle.\n\n**Turn count beats token price.** The cheapest model often takes 2-3× the turns on multi-step work — more wall-clock and context overall. Make a mid-tier model the floor for reviewers and for implementers working from prose; reserve the cheapest tier for transcription tasks whose exact code the plan already contains.\n\nTranslate the tier to a concrete model by judgment at dispatch time using your harness's available models. The three tiers are the whole ladder — `most-capable` is its top rung, not an open-ended \"best available\". A model that sits above that rung is outside the ladder: never reach it from a tier, and record a one-line reason in the ledger when you override to it. On a harness without per-dispatch model choice, everything runs on the session model — record that in the ledger.\n","\u003C!-- steepy:manual-handoff:v1:start -->\n## Manual handoff envelope\n\nManual phase boundaries use this literal shape (the envelope is chat text, never a file):\n\n```yaml\nhandoff: steepy-formula\u002Fv1\nnext: \u003Cphase>\nrequired: \u003Crole\u002Fpath map or none>\nonDemand: \u003Crole\u002Fpath map or none>\n```\n\nAll four keys are mandatory and occur once, in that order. Use explicit `none` for an empty map.\nThe phase input role maps are exact and unique: brainstorm has `required: none`, `onDemand: none`;\nplan has required `spec`, `onDemand: none`; Gear-2 direct implement has required `spec`,\n`onDemand: none`; plan-backed implement has required `plan`, on-demand `source-spec`; either\nimplement alternative additionally authorizes exact `progress-ledger` and `task-results` on demand\nwhen resuming;\nregular Gear-3 `review` has required `criteria`, `task-results`, and `branch-diff`, `onDemand: none`;\nfresh `loop-engineer` has required `goal`, `onDemand: none`; resume may additionally authorize exact\n`loop-ledger` on demand; Gear-4 `review` has required `goal` and `loop-ledger`, `onDemand: none`.\nThus Gear-2 brainstorm → implement maps `spec`; brainstorm → plan maps `spec`; plan → implement\nmaps `plan`, `source-spec`; Gear-3 implement → review maps `criteria`, `task-results`, `branch-diff`;\nloop-engineer → review maps `goal`, `loop-ledger`. The two implement role alternatives and the two\nreview role alternatives are mutually exclusive: never merge or infer them. A role occurs only in\nits mapped phase, alternative, and list.\n\nValidate literally before any work-artifact read. Reject duplicate keys, unknown keys, unknown roles,\nmissing keys, a wrong `next`, and unsafe paths: absolute paths, `..` segments, globs, or anything\noutside the repository. Never infer corrections. An explicit malformed or stale binding fails closed\nwithout discovery. Supporting `onDemand` inputs are never preloaded; read one only for a concrete\nnamed missing fact. Stable routing inputs stay outside the envelope and follow bootstrap routing.\nNative invocation spelling comes only from the active harness. No child inherits general\n`.formula\u002Fwork\u002F**` access; pass only its accepted role paths.\n\nEvery new work artifact starts with verdict\u002Fcontract metadata when that phase requires it, followed\nimmediately by this versioned header before the Markdown title:\n\n```yaml\n\u003C!-- steepy-workflow: v1\nphase: \u003Cbrainstorm|plan|implement|review|loop-engineer|goal-contract>\nstatus: \u003CDRAFT|READY|CONSUMED>\nnext: \u003Cphase|none>\nsource: \u003Csafe exact repo-relative path|none>\nconsumed-by: \u003Csafe exact repo-relative path|none>\n-->\n```\n\nHeader fields are mandatory, unique, and literal. `source` and `consumed-by` must agree with the\nadjacent lifecycle transition and authorize no transitive reads. Headerless artifacts are\nignored by recovery and are available only under explicit user audit scope.\n\n### Exact publication pairs\n\nThese are exact resume capabilities, not paths inferred from provenance. The input retains its\nnormal required role; the output role is added to `onDemand`, as are any existing on-demand companions.\nRegular review retains its required criteria and branch diff. Plan resume adds `onDemand.output-plan`;\nregular Gear-3 review resume adds `onDemand.review-report`. Fresh forms above are unchanged.\nGear-4 controller\u002Fledger handling is excluded from this model-based pair repair.\n\n| Route | Input role | Output role | Companion roles | Input phase | Output phase | Output next |\n| --- | --- | --- | --- | --- | --- | --- |\n| plan | spec | output-plan | none | brainstorm | plan | implement |\n| implement-2-direct | spec | task-results | progress-ledger | brainstorm | implement | none |\n| implement-2-plan | plan | task-results | source-spec,progress-ledger | plan | implement | none |\n| implement-3 | plan | task-results | source-spec,progress-ledger | plan | implement | review |\n| review-3 | task-results | review-report | criteria,branch-diff | implement | review | none |\n\n### Publication interruption prefixes\n\nThis table applies only to the exact authorized pair with matching phase, route and provenance.\nValidate the pair before the fresh-only READY input gate; a resume capability is never inferred.\nInput `next` names the consumer phase until consumed; consumption retains it for plan\u002Fimplement\nand sets it to `none` for review. Output `consumed-by` is `none` throughout this producer's run.\nAn unconsumed input must have `consumed-by: none`; a consumed input must name the exact output.\n\n| Input status | Output status | Action |\n| --- | --- | --- |\n| absent | absent | reject |\n| absent | DRAFT | reject |\n| absent | READY | reject |\n| absent | CONSUMED | reject |\n| DRAFT | absent | reject |\n| DRAFT | DRAFT | reject |\n| DRAFT | READY | reject |\n| DRAFT | CONSUMED | reject |\n| READY | absent | fresh-create |\n| READY | DRAFT | resume-work |\n| READY | READY | finish-consumption |\n| READY | CONSUMED | reject |\n| CONSUMED | absent | reject |\n| CONSUMED | DRAFT | reject |\n| CONSUMED | READY | no-op |\n| CONSUMED | CONSUMED | reject |\n\n`fresh-create` permits a new absent output only in a fresh run, never a missing resume target.\n`resume-work` requires exact output capability and leaves the input READY while work\u002Fgates remain.\nFor `finish-consumption`, revalidate output.source against the exact input path, input.consumed-by\nas `none`, and the recorded approval and verification bound to those input\u002Foutput paths and content\n(excluding lifecycle fields). Recheck current branch\u002FHEAD or uncommitted diff identity where applicable.\nOnly then perform the input-consumption write with input.consumed-by equal to the exact output path.\nFor `no-op`, revalidate the recorded proof with the final reciprocal links instead of the\nunconsumed-input condition; perform no writes or redispatch.\nA missing capability, mismatched path\u002Fphase\u002Fnext\u002Fcontent, absent approval\u002Fproof, or impossible state\nmust fail closed, preserving both files. A READY marker alone is not proof. Do not follow links to\nundeclared evidence; request the exact missing capability or user-delimited scope instead.\nNever select a most-recent output or recover by directory\u002Fbody search.\n\nPublication is ordered: record the phase's applicable approval and verification in the output,\npublish output READY by per-file atomic replacement, then consume the input by per-file atomic\nreplacement. This is not a multi-file atomic operation; READY\u002FREADY is its sole verified intermediate\nprefix. Resume repairs no other state and never repeats completed work or release actions.\nIn manual drive you are the enforcement: no parser, manifest, or validator checks this contract.\n\u003C!-- steepy:manual-handoff:v1:end -->\n\n\n# plan\n\nTurn an approved spec into an implementation plan whose every task states the change it makes, not\nthe goal it chases, with review checkpoints.\n\n> **Engine root:** this skill's base directory is `\u003Cengine-root>\u002Fskills\u002Fplan\u002F`; engine\n> scripts live two levels up, at `\u003Cengine-root>\u002Fscripts\u002F`. Resolve them relative to the base\n> directory your harness reports for this skill.\n\n## Checklist\n0. In autopilot, read the supplied phase manifest first; then read the gear and scale ceremony to it (Step 0).\n1. Load the spec and resolve the graph context (`context resolve`) (Step 1).\n2. Map files and tasks.\n3. Bind each task to the hub (surface, specialist agent, test command, complexity).\n4. Write the plan into `.formula\u002Fwork\u002Fplans\u002F`.\n4.5. Self-review the plan inline — free gate, gear 2 and 3 (Step 4.5).\n5. Validate with `steepy-formula graph validate` (deterministic gate).\n6. Human review gate — ask the user to review the written plan before handoff (gear 3).\n7. Hand off to the `implement` skill.\n\n## Procedure\n\n### Step 0 — Read the gear\n\n**Autopilot preflight:** When the conductor's `phasePrompt` supplies a phase manifest, read that\nmanifest before any other task input. Validate its role and scope against the plan phase and the\nsupplied correlation identity. Eagerly read every `required` input before acting. Do not preload `onDemand`;\nconsult it only for a concrete named missing fact and record the read in the plan phase\noutput\u002Freport. Phase-manifest authority applies only when `drive: autopilot`; manual-drive discovery\nremains Step 1's behavior.\n\n**Manual\u002Fno-manifest preflight — before the gear read:** Accept the exact `spec` capability before any verdict or spec-body read. Manual drive has two exclusive manual entry paths:\n\nFor an explicit resume, also require `onDemand.output-plan` in the envelope (or an exact labeled\n`output-plan` alongside the labeled spec). Validate that exact input\u002Foutput pair using the common\npublication table before the fresh-only READY input gate below: READY\u002FDRAFT resumes work,\nREADY\u002FREADY is finish-consumption, and CONSUMED\u002FREADY is no-op after proof revalidation. No other\nstate is accepted. The output header must be `phase: plan`, `next: implement`, with `source` equal\nto the accepted spec and `consumed-by: none`; a consumed spec must name this exact output.\nRead the output body only for the named missing approval\u002Fproof or draft-progress facts, record that\nreason there when resuming work, and do not reopen sibling work. A verified terminal pair goes\nstraight to Step 7 after any sole consumption repair, without writing another plan or repeating gates.\n\n1. **Explicit binding.** If the user supplied a `spec` handoff envelope or an exact spec path,\n   validate the literal binding and path safety before its header read. The path must be an exact\n   safe repository-relative path and the role must be exactly `spec`; then inspect only its bounded\n   `steepy-workflow` header and require `phase: brainstorm`, `status: READY`, and `next: plan`.\n   Exact valid input is accepted. Missing, malformed, stale, wrong-state, or wrong-role input fails\n   closed: report the literal defect and stop without discovery or inferred correction.\n2. **Bounded recovery.** Only when no envelope or path was supplied, enumerate the immediate specs\n   directory (`.formula\u002Fwork\u002Fspecs\u002F`, no recursion), inspect only bounded `steepy-workflow` headers,\n   list candidates whose header is exactly `phase: brainstorm`, `status: READY`, `next: plan` with\n   path and compact header metadata, and require user selection even when there is exactly one candidate.\n   Never open or read a candidate body. Headerless artifacts are ignored by\n   recovery. If no candidate exists, tell the user to run the `brainstorm` skill (invoke it the way\n   the active harness invokes skills) first and stop. After selection, obtain and validate the exact\n   binding before continuing.\n\nDo not read the selected spec's verdict or body until that exact binding is accepted.\nRead the verdict artifact (`verdict` + `gear`) only from the accepted `.formula\u002Fwork\u002F` spec head.\n- **Artifact present:** honor its gear.\n- **Artifact present with `gear: 4`:** this task belongs to the autonomous loop; bounce to\n  the `loop-engineer` skill and stop.\n- **Artifact with `drive: autopilot` (gear 3):** this run is an unattended autopilot phase. Confirm\n  that the contract agrees with the conductor-supplied phase manifest already read first. Apply the\n  Step 6 autopilot branch below and never ask the user anything.\n  An invalid manifest, missing required input, undeclared or unwritable output, or anything else\n  unresolvable is a manifest\u002Foutput failure: append `\u003CISO> — plan — BLOCKED — run-id=\u003Cconductor-supplied> attempt=\u003Cpositive supplied> \u003Creason>` to `.formula\u002Fwork\u002Ftasks\u002F\u003Cspec-basename>\u002Fautopilot-status.md`\n  and exit non-zero.\n- **Artifact absent and this is gear-1\u002F2 work:** recompute the coverage verdict inline (COVERED \u002F\n  CONFLICT \u002F GAP with `file:line` evidence → gear). Declare the recomputed\n  verdict, gear, and evidence, then ask the user for one confirmation before the ceremony starts —\n  the user may force a different gear, and the override is recorded in the verdict artifact. Then\n  proceed.\n- **Artifact absent and full gear-3 ceremony is required for this command:** do not proceed\n  blind — tell the user to enter through the upstream workflow chain first, and stop.\n\nScale this skill's ceremony to the gear (see \"Ceremony by gear\" below). This skill dispatches **no\nsubagent reviewer**: a plan is prose, gated by the inline self-review (Step 4.5) plus a human review\nof the written plan (Step 6), per the \"Review architecture (gear 3)\" governance rule —\nrecorded as a Rule in the canonical graph.\n\n**Autopilot correlation.** Before appending any current-protocol status marker, use the `run-id` and\npositive `attempt` supplied by the conductor's `phasePrompt`; that correlated identity is authoritative.\nDo not derive either value from filenames. Do not invent a correlation identity. A child without supplied run\u002Fattempt identity\nis unresolvable: exit non-zero without appending an uncorrelated new-protocol marker. Only versioned, correlated status is supported; uncorrelated markers never complete a phase.\n\n**Session declaration (declare, never block).** Say whether this conversation started clean for this\nphase or has already run an earlier phase of the chain in the same session. A same-session phase\nalready grew the context, and every turn of this phase now pays to carry it — name that cost, but\nnever refuse or gate on it: this is a declaration about the conversation, not a measurement, and\nsound work proceeds either way. If the session's history cannot be determined with certainty, say so\nand proceed. Under `drive: autopilot` this declaration is a no-op: the conductor\n(`scripts\u002Fautopilot.mjs`) already opens a fresh headless session per phase, so the discipline stated\nhere is already enforced by that mechanism, not by this prose.\n\n### Ceremony by gear\n\n- **Gear 1**: no plan needed — say so and stop.\n- **Gear 2**: an optional light plan is enough; self-review it (Step 4.5); skip only the human\n  question in Step 6, then perform the shared lifecycle transition.\n- **Gear 3**: run the full procedure, including the Step 4.5 self-review and the Step 6 human gate.\n- **Gear 4**: not this skill's entry point — bounce to the `loop-engineer` skill and stop.\n\n### Step 1 — Load the spec and the hub\n\n**Manual drive:** after Step 0 accepts the exact spec capability, continue with that path only.\nAfter stable bootstrap orientation, eagerly read only the accepted spec body, then resolve the\ngraph context. Never open a sibling or linked work artifact. The accepted `source`\u002F`consumed-by`\nfields authorize no transitive read.\n\n**Hub-mode declaration (once per run).** Run `steepy-formula hub-mode` and carry its declared\nmode — `graph`, `migrate`, or `none`, each printed with one `evidence:` line — through the whole\nrun. The command always exits 0: it is a declaration, not a gate, and this prose never\nre-implements its detection — the command is the single authority. The declared mode branches\nthe manual graph reads (autopilot's inventory is the phase manifest, below):\n\n- **`graph` mode:** resolve the hub context from the canonical graph — run\n  `steepy-formula context resolve --path \u003Cp>...` for the spec's exact paths (exit 0 emits\n  `context.json` + `context.md`). The resolved context supplies each owning **surface**, its\n  Rules — the standards-equivalent — its Terms, and each surface's test command from its Check\n  item's `ext_v1_command` extension property (fall back to the verification Rule's statement\n  prose when the Check carries no command). The bound specialist agent for surface `\u003Cname>` is\n  `.claude\u002Fagents\u002F\u003Cname>-agent.md` by naming convention. On exit 10 UNRESOLVED — normal in a freshly graph-first-initialized repo, where no\n  topology bridge links the detected Packages to the curated Components yet — fall back to\n  the deterministic surface targets:\n  `steepy-formula context resolve --target apex:\u003CprojectId>:surface:\u003Cslug(name)>` for each\n  owning surface, every part derived, never guessed — surface names from\n  `.claude\u002Fagents\u002F\u003Cname>-agent.md` (CLAUDE.md's Surfaces section is the fallback listing),\n  the project id from the hub-mode declaration's `hub-mode '\u003CprojectId>'` echo, `slug(name)`\n  lowercase with non-alphanumeric runs collapsed to `-` — and never target the bare project\n  root on its own: it resolves empty. If the resolve fails, diagnose with\n  `steepy-formula graph status` — exit 3,\n  `initialized: no`, or an engine failure means the graph binding is missing or failed: tell the\n  user to run the `init` skill first and stop.\n- **`migrate` mode:** the repo still carries a pre-graph Markdown hub.\n  Stop and tell the user to run the `graph-migrate` skill (invoke it the way your harness invokes skills) before any other workflow.\n- **Mode `none`:** no graph — the repo is not initialized: tell the user to run\n  the `init` skill first and stop.\n\n**Autopilot:** use the validated phase manifest loaded in Step 0 as the authoritative input\ninventory. Do not independently rediscover or preload inputs beyond its `required` list; use the\nrecorded `onDemand` protocol only when a concrete fact is missing.\n\n### Step 2 — Map files and tasks\n\nList the files each task creates\u002Fmodifies (exact paths). A task is mis-cut if its implementer must **discover** what to change: when its requirements can be stated only as a goal (\"make X work,\" \"write the section on Y\") and not as a sketch of a diff to exact paths, the discovery is missing and belongs to `plan` — split the task, or add a discovery task upfront whose deliverable is the missing facts. This is the other half of what a task already needs: not just the exact paths, but the expected change at each path. Decompose into tasks, each ending in an independently testable deliverable.\n\n**Wide refactors — expand → migrate → contract.** A **wide refactor** — one mechanical change\nwhose blast radius breaks callers across the whole codebase — must not be forced into tasks\nthat cannot land green. Sequence it: **expand** (add the new form beside the old — nothing\nbreaks), **migrate** (move call sites in batches sized by blast radius; each batch is an\nordinary sequential task, suite green between batches), **contract** (delete the old form when\nno caller remains). No dependency-graph machinery — batches are ordinary sequential plan\ntasks. \"Wide\" is strict: if the change lands green in one task, it is not wide.\n\n### Step 3 — Bind each task to the hub\n\nThe plan is the complete spec-to-brief bridge. Every task uses this self-contained format:\n\nStart every task with the canonical boundary `## Task \u003Cpositive integer> — \u003Cshort title>`; task\nintegers are unique and increase in execution order. Under it use the exact labels below. The\ncanonical criterion form is `- **Success criteria:** SC1, SC2`; the parser rejects alternate field names such as `Spec criteria`.\n\n- **Requirements and deliverables:** exact behavior to implement and the independently testable result.\n- **Relevant global constraints:** every spec-wide rule that applies to this task, repeated here rather than assumed.\n- **Surface:** `\u003Csurface>` — from the resolved context.\n- **Specialist agent:** the `\u003Csurface>-agent` that owns implementation — bound as `.claude\u002Fagents\u002F\u003Csurface>-agent.md` by naming convention.\n- **Exact paths:** every created or modified repository-relative path; paths must be safe and exact.\n- **Test command:** the surface's exact command — the resolved context Check's `ext_v1_command` (falling back to the verification Rule's statement prose).\n- **Dependencies:** prior task numbers, or `none`, in executable order.\n- **Complexity:** `mechanical | integration | design` — mechanical = 1-2 files with a complete\n  spec; integration = multi-file or judgment; design = architecture, high-risk, or subtle.\n- **Success criteria:** every mapped success-criterion ID from the source spec.\n\nInclude all exact requirements and deliverables needed to materialize the downstream task brief.\nNo requirement may live only in conversation. If copying a large upstream section would make the\ntask unwieldy, reference it by canonical path + heading and state exactly which requirement it\nsupplies. `implement` reads this line (`Complexity`) to pick the model tier and decide whether the task gets its\nown reviewer (Step 3.4).\n\n### Step 4 — Write the local working plan\n\n**Manual drive:** on a fresh run, choose an absent `.formula\u002Fwork\u002Fplans\u002FYYYY-MM-DD-\u003Ctopic>.md` path;\nan existing target requires the exact resume capability before reading or overwriting it. On resume,\nretain the accepted `output-plan` path, never recalculate its date. After any verdict\u002Fcontract metadata and before the title, write:\n\n```yaml\n\u003C!-- steepy-workflow: v1\nphase: plan\nstatus: DRAFT\nnext: implement\nsource: \u003Cexact-spec-path>\nconsumed-by: none\n-->\n```\n\nThen link back to its source spec with a relative work-area link:\n\n`> Source spec: [\u003Ctopic>](..\u002Fspecs\u002F\u003CYYYY-MM-DD-topic>.md)`\n\n**Autopilot:** write the plan exactly to the single path declared by the authoritative phase\nmanifest's `outputs` list. Do not recompute the date or filename. Use that actual path in the\ncorrelated `DONE` marker, and derive the relative source-spec back-link from its location.\n\nCreate `.formula\u002Fwork\u002Fplans\u002F` if it does not exist. Do not create a plans sub-index. Plans are local workflow artifacts, not stable hub documentation.\n\n### Step 4.5 — Self-review the plan (free gate)\n\n**Gear 2 and 3** — skip only at gear 1 (no plan is produced there).\n\nRead the plan you just wrote with fresh eyes and fix any of these inline — no re-review needed, just fix and move on:\n1. **Every mandatory field** — each task contains exact requirements\u002Fdeliverables, relevant global constraints, surface, specialist agent, exact paths, test command, dependencies, complexity, and success-criterion IDs; no placeholder (`TBD`, `TODO`) remains.\n2. **Binding consistency** — every surface is a known surface among the resolved context's Surfaces, every specialist is `.claude\u002Fagents\u002F\u003Csurface>-agent.md` by naming convention, and every exact test command matches the Check's `ext_v1_command` (or its verification Rule's statement fallback).\n3. **Dependency ordering** — dependencies name existing earlier tasks, ordering is executable, and each task ends in an independently testable deliverable.\n4. **Complexity present** — validate allowed complexity: each `Complexity` value is exactly `mechanical | integration | design` and matches the task text (a multi-file task marked `mechanical` is a bug).\n5. **Criteria coverage** — every requirement and success criterion in the spec maps to at least one task; list and close any gap.\n6. **Safe repository-relative paths** — every path is exact, stays inside the repository, and is owned by the task's declared surface.\n7. **Context independence** — a downstream agent can implement the task from the plan\u002Ftask brief without chat history; large references use canonical path + heading and state the supplied requirement.\n8. **Discovery-free tasks** — no task's requirements are stated only as a goal (\"make X work,\" \"write the section on Y\"); each names exact paths and the expected change at each path, or is split \u002F preceded by a discovery task whose deliverable is the missing facts.\n\nAll checks must pass before appending autopilot `DONE` or asking for human approval.\n\n### Step 5 — Validate (deterministic gate)\n\nRun the read-only graph gate:\n\n```bash\nsteepy-formula graph validate\n```\n\nIt MUST exit 0 with `violations: 0` and a clean revision-chain replay. This skill never edits the graph, so any violation is pre-existing drift: report it and point the user to the `check` skill — do not fix it mid-plan.\n\n**Autopilot:** Before appending `DONE`, run the same canonical plan parse that the implement\nphase will use:\n\n```bash\nnode \u003Cengine-root>\u002Fscripts\u002Fautopilot-context.mjs --verify-plan --repo-root . --plan \u003Cauthoritative-manifest-output-path>\n```\n\nIt MUST print `plan OK`. A rejection is a fixable plan error: repair the plan, repeat the Step 4.5\nself-review, and rerun both deterministic gates. Never append `DONE` for a plan this command has\nnot accepted.\n\n### Step 6 — Human review gate and plan lifecycle\n\n**Gear 3 only — human review gate.** A plan is prose. Every task it defines will pass under the per-task and\nwhole-branch code reviewers downstream, but the plan's *shape* — decomposition, spec coverage, task\nordering — is best caught by a human now. Do **not** dispatch a subagent. Tell the user:\n\n> \"Plan written to `\u003Cpath>` — please review it and tell me if anything should change before we implement.\"\n\nWait for the response. On requested changes, edit the plan and re-run the Step 4.5 self-review.\nProceed only once the user approves.\n\n**Gear 2 gate.** Ask no human-review question; the completed Step 4.5 self-review is the applicable\napproval gate.\n\n**Both gears.** Record the applicable approval and verification (Step 4.5, Step 5, and the human\napproval or autopilot authorization) bound to the exact source\u002Foutput paths and content before READY.\nOnly after those gates succeed, first atomically replace the plan to `status: READY` while retaining\n`source: \u003Cexact-spec-path>`; then atomically replace the source spec to `status: CONSUMED` with\n`consumed-by: \u003Cplan-path>`. Before publication, interruption leaves the spec READY and plan DRAFT;\nbetween the two replacements, READY\u002FREADY is the sole repairable prefix. Use only the exact resume\ncapabilities and common proof checks to finish consumption; a finalized pair is an idempotent no-op.\nNever repair a CONSUMED-spec\u002FDRAFT-plan pair or consume an unverified plan.\n\nAutopilot status timestamps must use UTC with three millisecond digits (`YYYY-MM-DDTHH:mm:ss.sssZ`), generated with `new Date().toISOString()`.\n\n**Autopilot:** If `drive: autopilot` is present in the verdict contract (Step 0), the human gate is superseded — this is a non-blocking checkpoint. The plan file is on disk; the human can interrupt the run. Skip the question, do NOT wait, and append `\u003CISO timestamp> — plan — DONE — run-id=\u003Cconductor-supplied> attempt=\u003Cpositive supplied> \u003Cplan path>` to `.formula\u002Fwork\u002Ftasks\u002F\u003Cspec-basename>\u002Fautopilot-status.md` (spec filename without `.md`; the status file already exists). Step 4.5 self-review and Step 5 `graph validate` still run and still gate.\n\n### Step 7 — Hand off\n\nOnly after the applicable Gear-2 or Gear-3 gate and lifecycle transition succeed, tell the user the next step is the `implement` skill, using the\nnative invocation spelling only when the active harness provides it. In manual drive emit:\n\n```yaml\nhandoff: steepy-formula\u002Fv1\nnext: implement\nrequired:\n  plan: \u003Cplan-path>\nonDemand:\n  source-spec: \u003Cspec-path>\n```\n\nRun the `implement` skill in a new session: this phase's context does not serve implementation, and every turn of the implement phase would pay to carry it forward.\n\n## Model Selection\n\nAlways set `model:` explicitly when dispatching a subagent. An omitted model inherits the session model (usually the most capable and most expensive) and silently defeats this policy.\n\nPick the tier by task complexity and risk:\n\n| Signal | Tier |\n|--------|------|\n| Mechanical \u002F transcription (1-2 files, complete spec) | cheap |\n| Integration \u002F multi-file \u002F judgment | standard |\n| Design \u002F architecture \u002F high-risk or subtle change | most-capable |\n\nReviewers floor at **standard** and rise to **most-capable** when the artifact is high-risk or subtle.\n\n**Turn count beats token price.** The cheapest model often takes 2-3× the turns on multi-step work — more wall-clock and context overall. Make a mid-tier model the floor for reviewers and for implementers working from prose; reserve the cheapest tier for transcription tasks whose exact code the plan already contains.\n\nTranslate the tier to a concrete model by judgment at dispatch time using your harness's available models. The three tiers are the whole ladder — `most-capable` is its top rung, not an open-ended \"best available\". A model that sits above that rung is outside the ladder: never reach it from a tier, and record a one-line reason in the ledger when you override to it. On a harness without per-dispatch model choice, everything runs on the session model — record that in the ledger.\n",{"name":5,"description":7,"user-invocable":24},true,"main","a5e0492607593191c2b0a1c603cf421996b5e22f","skills\u002Fplan",0,"https:\u002F\u002Fgithub.com\u002Fsteepy-it\u002Fsteepy-formula","https:\u002F\u002Fwww.steepy.it","https:\u002F\u002Fgithub.com\u002Fsteepy-it","2026-10-27 16:48:08",{"source":8,"skillName":5,"commands":34,"prompt":44,"zipAvailable":24},[35,38,41],{"tool":36,"command":37},"npx","npx skills add steepy-it\u002Fsteepy-formula --skill plan",{"tool":39,"command":40},"bunx","bunx skills add steepy-it\u002Fsteepy-formula --skill plan",{"tool":42,"command":43},"pnpm","pnpm dlx skills add steepy-it\u002Fsteepy-formula --skill plan","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Fsteepy-it\u002Fsteepy-formula\u002Fblob\u002Fmain\u002Fskills\u002Fplan\u002FSKILL.md",[],[47,54,61,68,75,82],{"skillNo":48,"name":49,"nameZh":6,"description":50,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":51,"categories":52,"keywords":53,"publishedAt":19},"01M7F91BV8VJM3DS4SA9ZMHMXB","web-check","Run the web surface checks.","https:\u002F\u002Fgithub.com\u002Fsteepy-it\u002Fsteepy-formula\u002Fblob\u002Fmain\u002Ftests\u002Ffixtures\u002Fgood-hub\u002F.claude\u002Fskills\u002Fweb-check\u002FSKILL.md",[],[],{"skillNo":55,"name":56,"nameZh":6,"description":57,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":58,"categories":59,"keywords":60,"publishedAt":19},"01M7F917VTJWS90304QDH1C3R8","review","Verify implemented work against its spec with real evidence before any \"done\" claim. Checks the spec's success criteria, runs the surface test command, and runs `steepy-formula graph validate` through the deterministic inline evidence capture.","https:\u002F\u002Fgithub.com\u002Fsteepy-it\u002Fsteepy-formula\u002Fblob\u002Fmain\u002Fskills\u002Freview\u002FSKILL.md",[],[],{"skillNo":62,"name":63,"nameZh":6,"description":64,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":65,"categories":66,"keywords":67,"publishedAt":19},"01M7F90WV19MCWWJFESM2BNHNN","new-surface","Add a surface: render the graph sub-agent and promote Surface + verification Rule + Check into the graph.","https:\u002F\u002Fgithub.com\u002Fsteepy-it\u002Fsteepy-formula\u002Fblob\u002Fmain\u002Fskills\u002Fnew-surface\u002FSKILL.md",[],[],{"skillNo":69,"name":70,"nameZh":6,"description":71,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":72,"categories":73,"keywords":74,"publishedAt":19},"01M7F90QA1DXYPVFD08V5HFYY1","loop-engineer","Validate and authorize a gear-4 goal contract, invoke the packaged deterministic loop controller, report its result, and hand off terminal evidence to review.","https:\u002F\u002Fgithub.com\u002Fsteepy-it\u002Fsteepy-formula\u002Fblob\u002Fmain\u002Fskills\u002Floop-engineer\u002FSKILL.md",[],[],{"skillNo":76,"name":77,"nameZh":6,"description":78,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":79,"categories":80,"keywords":81,"publishedAt":19},"01M7F90EZHJZKYXEWTJ3BWKWRK","init","Graph-first init — gate on a provisioned Stable Knowledge Graph before writing anything, interview the governed structure, scaffold the disk-minimal harness assets, and land the initial graph through a promotion scenario. Never invents Rules from stack detection.","https:\u002F\u002Fgithub.com\u002Fsteepy-it\u002Fsteepy-formula\u002Fblob\u002Fmain\u002Fskills\u002Finit\u002FSKILL.md",[],[],{"skillNo":83,"name":84,"nameZh":6,"description":85,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":86,"categories":87,"keywords":88,"publishedAt":19},"01M7F908EYQDZ71GA3TQZ0613E","inception","Turn an idea and its starting materials into a minimal, verified, runnable application before the graph binding exists — reconnaissance, architecture decisions, verified versions, one human approval, bootstrap of one representative path, and final verification — then hand off to init and discovery.","https:\u002F\u002Fgithub.com\u002Fsteepy-it\u002Fsteepy-formula\u002Fblob\u002Fmain\u002Fskills\u002Finception\u002FSKILL.md",[],[]]