[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M7CX0G1MHSJEXCBY0PBQ05Z2":3,"skill-same:01M7CX0G1MHSJEXCBY0PBQ05Z2":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":24,"commitSha":25,"skillDirectory":26,"forksCount":27,"repositoryHtmlUrl":28,"repositoryHomepageUrl":28,"ownerHtmlUrl":29,"ownerBlogUrl":6,"ownerDescription":30,"updatedAt":31,"install":32,"resources":45},"01M7CX0G1MHSJEXCBY0PBQ05Z2","auto-rolling-roadmap",null,"Maintain ROADMAP.md as the project's living product, architecture, and long-horizon development checkpoint. Use when steering project direction, updating roadmap checkpoints, or resolving code soup drift.","CardSorting\u002FHEAV3NS","CardSorting","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F115497755?v=4","01M7AFCH0P4ZXA11RX8SVJY25G","HEAV3NS","01M7CWZ7ZHR0T7NG9QAMKS1XAQ",1,"Apache-2.0","https:\u002F\u002Fgithub.com\u002FCardSorting\u002FHEAV3NS\u002Fblob\u002Fmain\u002Foptional-skills\u002Fdietcode\u002Fauto-rolling-roadmap\u002FSKILL.md",[],[],"2026-11-15 14:19:17",false,"---\nname: auto-rolling-roadmap\ndescription: Maintain ROADMAP.md as the project's living product, architecture, and long-horizon development checkpoint. Use when steering project direction, updating roadmap checkpoints, or resolving code soup drift.\n---\n\n# DietCode Skill: Auto-Rolling Roadmap Checkpoint System\n\n## Skill Purpose\n\nMaintain an evolving `ROADMAP.md` file that acts as the project's living product, architecture, and long-horizon development checkpoint.\n\nThis skill exists to prevent code soup, preserve project coherence, and keep long-running development aligned around a clear center of gravity.\n\nThe roadmap is not a wishlist.\n\nThe roadmap is not a backlog.\n\nThe roadmap is not a promise.\n\nThe roadmap is the project's steering surface.\n\nIt should help humans and agents answer:\n\n* What is this project becoming?\n* What matters now?\n* What should happen next?\n* What should remain deferred?\n* What changed recently?\n* What risks are accumulating?\n* Is the system becoming more coherent or more fragmented?\n\n## Primary Output\n\nThis skill must create or update:\n\n```text\nROADMAP.md\n```\n\n`ROADMAP.md` is the canonical checkpoint file for long-horizon project steering.\n\nIf the file does not exist, create it.\n\nIf the file exists, evolve it carefully.\n\nDo not replace useful history with generic text.\n\nDo not append endlessly.\n\nDo not turn the roadmap into a backlog dump.\n\nEach update should make the roadmap more useful, more current, and easier to navigate.\n\n## Prime Directive\n\nEvery roadmap pass must answer:\n\n```text\nDid the latest work strengthen or weaken the project's center of gravity?\n```\n\nIf the answer is unclear, the skill must investigate before adding or promoting roadmap items.\n\nA roadmap item that does not connect to the project's center of gravity should remain in Discovery, be reframed, or be archived.\n\n## Definition: Center of Gravity\n\nThe project's center of gravity is the smallest set of concepts, workflows, abstractions, and operational paths that explain how the system fundamentally works.\n\nA strong center of gravity means:\n\n* there is one obvious way to perform core operations\n* agents can predict where changes belong\n* contributors can understand the system without reading everything\n* new features attach to existing architecture instead of creating parallel systems\n* operational behavior is inspectable\n* maintenance burden stays bounded\n* long-term development feels navigable\n\nA weak center of gravity means:\n\n* duplicate abstractions appear\n* multiple workflows compete\n* state ownership becomes unclear\n* debugging paths fragment\n* agents invent local patterns\n* hidden orchestration grows\n* roadmap direction becomes disconnected from implementation reality\n\n## Core Philosophy\n\nCentralization does not mean giant files, giant classes, or rigid architecture.\n\nCentralization means clear authority.\n\nModules may be distributed.\n\nAuthority should not be.\n\nThe roadmap should continuously protect:\n\n* canonical workflows\n* canonical mutation paths\n* canonical runtime inspection\n* canonical state ownership\n* canonical terminology\n* canonical architectural intent\n\nThe system should feel extensible without becoming incoherent.\n\n## Required Inputs\n\nWhen invoked, inspect available project context before editing `ROADMAP.md`.\n\nUse any available sources such as:\n\n* existing `ROADMAP.md`\n* README files\n* architecture docs\n* recent commits\n* changed files\n* tests\n* scripts\n* issue notes\n* implementation summaries\n* failing commands\n* current user request\n* agent output logs\n* TODOs or FIXME markers\n* package\u002Fconfig changes\n\nDo not rely only on the user's latest request if project artifacts are available.\n\nDo not invent project state.\n\nIf evidence is missing, mark uncertainty explicitly.\n\n## Per-Project Identity (DietCode native integration)\n\nEvery `roadmap` tool response is scoped to the **Hermes project workspace**, not\nthe DietCode plugin install tree. Read these fields before editing:\n\n| Field | Use |\n| --- | --- |\n| `project_identity_line` | One-line header: project brief · stack · verify command |\n| `project_steering_digest` | Entity card: CI, quality tools, governance, verify, bootstrap status |\n| `project_fingerprint` | Raw signals in checkpoint `evidence` (README title, archetype, Makefile targets, …) |\n| `bootstrap_fill_plan` | When template phrases remain — use `tasks[].suggested_replacement` |\n\nPrefer evidence-backed replacements over generic text. When placeholders remain,\ncall `roadmap(action='apply_bootstrap_fill')` before manual edits.\n\nOperator reference: `docs\u002Froadmap.md` in the DietCode plugin tree.\n\nKey doc sections for agents:\n\n- **12-section schema** — required headings and enumerated health\u002Fsoup values\n- **Example payloads** — shape of `guide`, `checkpoint`, and `validate` responses\n- **Bootstrap fill** — `bootstrap_fill_plan.tasks[].suggested_replacement`\n- **Write guard** — ROADMAP.md only at `{workspace}\u002FROADMAP.md`\n- **Anti-patterns** — backlog dumping, skipping §9 audit, plugin-tree writes\n\n## Required Update Algorithm\n\nFollow this sequence on every roadmap pass:\n\n1. Read the existing `ROADMAP.md`, if present.\n2. Identify the current stated center of gravity.\n3. Inspect recent project changes and available evidence.\n4. Determine whether the project is coherent, accelerating, drifting, fragmenting, blocked, or overloaded.\n5. Compare current work against the existing roadmap.\n6. Preserve valid strategic intent.\n7. Remove duplicate or stale roadmap entries.\n8. Archive items that no longer deserve active attention.\n9. Promote items only when evidence supports promotion.\n10. Demote items when uncertainty, risk, or entropy increases.\n11. Add new items only when they connect to the center of gravity.\n12. Rewrite technical implementation details into clear product and architecture language.\n13. Run a centralization and code soup audit.\n14. Update the recent checkpoint.\n15. Add decision log entries for meaningful direction changes.\n16. Return a concise summary of what changed.\n\n## ROADMAP.md Required Structure\n\nThe file should use this structure.\n\n```markdown\n# ROADMAP.md\n\n## 1. Project Center of Gravity\n\n## 2. Roadmap Health\n\n## 3. Strategic Narrative\n\n## 4. Now\n\n## 5. Next\n\n## 6. Later\n\n## 7. Discovery\n\n## 8. Maintenance Gravity\n\n## 9. Centralization & Code Soup Audit\n\n## 10. Decision Log\n\n## 11. Recent Checkpoint\n\n## 12. Archive\n```\n\nDo not rename these top-level sections unless the user explicitly requests a different schema.\n\nThe consistency of the schema is part of the skill's value.\n\n## Section Contract\n\n### 1. Project Center of Gravity\n\nDescribe what the project fundamentally is.\n\nThis section should be understandable to a non-technical stakeholder while still useful to an engineer or agent.\n\nInclude:\n\n```markdown\n## 1. Project Center of Gravity\n\n**Core Purpose:**  \n\u003Cplain-language purpose>\n\n**Primary Users \u002F Operators:**  \n\u003Cwho uses or operates this system>\n\n**Canonical Architecture:**  \n\u003Cshort description of the main architectural shape>\n\n**Canonical Workflows:**  \n\u003Cthe main flows agents and humans should preserve>\n\n**Primary Runtime \u002F Operational Center:**  \n\u003Cwhere operational truth lives>\n\n**What This Project Must Not Become:**  \n\u003Canti-goals that protect coherence>\n```\n\nThe \"must not become\" field is mandatory.\n\nIt prevents drift.\n\nExamples:\n\n* must not become a scattered collection of scripts\n* must not become another generic AI IDE\n* must not create multiple competing orchestration systems\n* must not hide mutations behind invisible automation\n* must not trade inspectability for convenience\n\n### 2. Roadmap Health\n\nGive the current health state.\n\nAllowed statuses:\n\n```text\nCoherent\nAccelerating\nDrifting\nFragmenting\nBlocked\nOverloaded\nRecovering\n```\n\nUse exactly one primary status.\n\nFormat:\n\n```markdown\n## 2. Roadmap Health\n\n**Status:** Coherent \u002F Accelerating \u002F Drifting \u002F Fragmenting \u002F Blocked \u002F Overloaded \u002F Recovering\n\n**Summary:**  \n\u003Cbrief explanation>\n\n**Why This Status:**  \n- \u003Cevidence>\n- \u003Cevidence>\n- \u003Cevidence>\n\n**Primary Risk:**  \n\u003Csingle biggest risk>\n\n**Primary Opportunity:**  \n\u003Csingle biggest opportunity>\n```\n\n### 3. Strategic Narrative\n\nExplain what the project is becoming.\n\nThis should be short, clear, and directional.\n\nAvoid vague hype.\n\nFormat:\n\n```markdown\n## 3. Strategic Narrative\n\n\u003C2-5 paragraphs explaining the current product \u002F architecture direction>\n\nThe narrative should clarify:\n- what changed recently\n- what the system is converging toward\n- what should be protected\n- what should be avoided\n```\n\n### 4. Now\n\nActive, committed, or already-in-motion work.\n\nOnly include items that are ready for execution or currently underway.\n\nDo not include vague ideas.\n\nEach item must use this format:\n\n```markdown\n### N. \u003CItem Title>\n\n**Goal:**  \n\u003Cwhat this work accomplishes>\n\n**Why It Matters:**  \n\u003Cproduct, architecture, user, or maintenance value>\n\n**Current State:**  \n\u003Cknown state from evidence>\n\n**Next Concrete Action:**  \n\u003Cone specific next action>\n\n**Success Signal:**  \n\u003Chow we know this worked>\n\n**Gravity Impact:** Strengthens \u002F Neutral \u002F Weakens \u002F Unknown  \n**Centralization Effect:** Centralizes \u002F No Change \u002F Decentralizes  \n**Entropy Risk:** Low \u002F Medium \u002F High  \n\n**Risk \u002F Mitigation:**  \n\u003Crisk and mitigation>\n```\n\nRules for Now:\n\n* keep it small\n* prefer 1-5 items\n* each item must be actionable\n* each item must have evidence\n* each item must connect to the center of gravity\n\nIf there are more than 5 Now items, the roadmap is overloaded.\n\n### 5. Next\n\nLikely upcoming work.\n\nThese are prepared but not fully committed.\n\nFormat:\n\n```markdown\n### N. \u003CItem Title>\n\n**Opportunity:**  \n\u003Cwhat could be unlocked>\n\n**Why Soon:**  \n\u003Cwhy this may matter after Now>\n\n**Dependency:**  \n\u003Cwhat must happen first>\n\n**First Validation Step:**  \n\u003Csmallest way to test whether this deserves promotion>\n\n**Confidence:** High \u002F Medium \u002F Low  \n\n**Gravity Impact:** Strengthens \u002F Neutral \u002F Weakens \u002F Unknown  \n**Centralization Effect:** Centralizes \u002F No Change \u002F Decentralizes  \n**Entropy Risk:** Low \u002F Medium \u002F High\n```\n\nRules for Next:\n\n* do not pretend Next is committed\n* include validation steps\n* use confidence labels\n* avoid implementation rabbit holes\n\n### 6. Later\n\nLong-horizon possibilities.\n\nThese are strategic directions, not promises.\n\nFormat:\n\n```markdown\n### N. \u003CDirection Title>\n\n**Direction:**  \n\u003Cwhat this could become>\n\n**Potential Upside:**  \n\u003Cwhy it may matter>\n\n**Why Not Now:**  \n\u003Cwhy it remains deferred>\n\n**Promotion Trigger:**  \n\u003Cwhat evidence would move this into Next>\n\n**Entropy Risk:** Low \u002F Medium \u002F High\n```\n\nRules for Later:\n\n* keep speculative work out of Now\n* preserve interesting possibilities without letting them hijack execution\n* include promotion triggers so Later does not become a junk drawer\n\n### 7. Discovery\n\nQuestions, experiments, unknowns, and ambiguous ideas.\n\nDiscovery exists to protect the roadmap from false certainty.\n\nFormat:\n\n```markdown\n### N. \u003CQuestion or Investigation>\n\n**Question:**  \n\u003Cwhat needs to be understood>\n\n**Evidence Needed:**  \n\u003Cwhat would clarify this>\n\n**Possible Outcomes:**  \n- \u003Coutcome>\n- \u003Coutcome>\n- \u003Coutcome>\n\n**Decision Needed:**  \n\u003Cwhat decision this discovery should eventually support>\n```\n\nRules for Discovery:\n\n* vague ideas belong here, not Now\n* technical uncertainty belongs here\n* product uncertainty belongs here\n* architectural ambiguity belongs here\n\n### 8. Maintenance Gravity\n\nTrack places where complexity, confusion, or fragility is accumulating.\n\nFormat:\n\n```markdown\n## 8. Maintenance Gravity\n\n### Hotspots\n\n| Area | Symptom | Risk | Recommended Action |\n|---|---|---|---|\n| \u003Carea> | \u003Csymptom> | Low\u002FMedium\u002FHigh | \u003Caction> |\n\n### Repeated Friction\n\n- \u003Cbug, workflow issue, agent confusion, or maintenance pattern>\n\n### Documentation Gaps\n\n- \u003Cmissing or stale docs>\n\n### Agent Confusion Points\n\n- \u003Cplaces agents are likely to patch incorrectly>\n```\n\nMaintenance gravity is not shame.\n\nIt is early warning.\n\nThe purpose is to detect entropy before it becomes structural collapse.\n\n### 9. Centralization & Code Soup Audit\n\nEvaluate whether the project is becoming more or less coherent.\n\nUse this checklist:\n\n```markdown\n## 9. Centralization & Code Soup Audit\n\n**Overall Code Soup Risk:** Low \u002F Medium \u002F High\n\n### Canonical Path Integrity\n\n- Are there multiple ways to perform the same operation?\n- Are there duplicate APIs, scripts, commands, or workflows?\n- Is the correct path obvious?\n\n**Assessment:**  \n\u003Canswer>\n\n### Authority Boundaries\n\n- Where does runtime authority live?\n- Where does state authority live?\n- Where does mutation authority live?\n- Where does diagnostic authority live?\n\n**Assessment:**  \n\u003Canswer>\n\n### Structural Drift\n\n- Did recent work introduce isolated patterns?\n- Did it create one-off abstractions?\n- Did it bypass existing architecture?\n\n**Assessment:**  \n\u003Canswer>\n\n### Agent Coherence\n\n- Would a future agent know where to patch?\n- Would a future agent reuse the canonical path?\n- Did recent work make agent behavior safer or more chaotic?\n\n**Assessment:**  \n\u003Canswer>\n\n### Centralization Recommendation\n\n\u003Cone recommendation to strengthen project gravity>\n```\n\nThis section is mandatory.\n\nIf the roadmap does not audit code soup risk, the skill failed.\n\n### 10. Decision Log\n\nRecord meaningful product, architecture, and workflow decisions.\n\nFormat:\n\n```markdown\n## 10. Decision Log\n\n### YYYY-MM-DD — \u003CDecision Title>\n\n**Decision:**  \n\u003Cwhat was decided>\n\n**Reason:**  \n\u003Cwhy>\n\n**Impact:**  \n\u003Cwhat changes because of this>\n\n**Follow-up:**  \n\u003Cwhat should happen next>\n```\n\nRules:\n\n* add entries only for meaningful decisions\n* do not log every tiny change\n* do not rewrite history casually\n* preserve old decisions unless they are clearly obsolete\n* if a decision is reversed, add a new entry instead of deleting the old one\n\n### 11. Recent Checkpoint\n\nSummarize the latest roadmap update.\n\nFormat:\n\n```markdown\n## 11. Recent Checkpoint\n\n**Date:** YYYY-MM-DD\n\n**Checkpoint Summary:**  \n\u003Cwhat changed this pass>\n\n**Moved:**  \n- \u003Citem moved from one section to another>\n\n**Added:**  \n- \u003Cnew item>\n\n**Updated:**  \n- \u003Cexisting item updated>\n\n**Archived:**  \n- \u003Carchived item>\n\n**Code Soup Risk:** Low \u002F Medium \u002F High  \n\u003Cbrief reason>\n\n**Recommended Next Move:**  \n\u003Cone concrete action>\n```\n\nOnly one Recent Checkpoint should exist.\n\nWhen updating, replace the previous Recent Checkpoint with the latest one.\n\nOlder meaningful changes should be preserved in the Decision Log or Archive.\n\n### 12. Archive\n\nThe archive stores removed, deferred, killed, or superseded roadmap items.\n\nFormat:\n\n```markdown\n## 12. Archive\n\n### \u003CArchived Item>\n\n**Archived Date:** YYYY-MM-DD  \n**Reason:**  \n\u003Cwhy it was removed from active attention>\n\n**Restore Condition:**  \n\u003Cwhat would justify bringing it back>\n```\n\nRules:\n\n* archive instead of deleting when strategic memory matters\n* delete only if the item was duplicate, noise, or clearly useless\n* archive stale speculative work\n* archive work that weakens center of gravity without sufficient upside\n\n## Roadmap Item Scoring\n\nEvery Now, Next, and Later item must include gravity metadata.\n\nUse these exact labels.\n\n### Gravity Impact\n\n```text\nStrengthens\nNeutral\nWeakens\nUnknown\n```\n\nMeaning:\n\n* Strengthens: reinforces the project center of gravity\n* Neutral: does not meaningfully affect coherence\n* Weakens: fragments authority, duplicates systems, or increases drift\n* Unknown: not enough evidence yet\n\n### Centralization Effect\n\n```text\nCentralizes\nNo Change\nDecentralizes\n```\n\nMeaning:\n\n* Centralizes: moves authority, diagnostics, workflow, or state closer to canonical surfaces\n* No Change: does not affect centralization\n* Decentralizes: creates new authority surfaces, parallel paths, or hidden behavior\n\n### Entropy Risk\n\n```text\nLow\nMedium\nHigh\n```\n\nMeaning:\n\n* Low: unlikely to create maintenance burden\n* Medium: manageable but requires attention\n* High: likely to create fragmentation, confusion, or long-term maintenance cost\n\n## Promotion and Demotion Rules\n\n### Promote Discovery to Next only when:\n\n* the question is clear\n* evidence exists\n* the opportunity connects to the center of gravity\n* the first validation step is known\n* entropy risk is understood\n\n### Promote Next to Now only when:\n\n* dependencies are resolved\n* the next action is concrete\n* success signal is clear\n* the work is aligned with current strategy\n* the roadmap has capacity\n\n### Demote Now to Next when:\n\n* work is blocked\n* confidence drops\n* dependencies are missing\n* implementation risk grows\n* it no longer fits the current center of gravity\n\n### Move anything to Archive when:\n\n* it is stale\n* it duplicates another item\n* it weakens project gravity\n* it belongs to an older strategy\n* it is not worth preserving in active attention\n\n## Anti-Bloat Rules\n\nThe roadmap must evolve by compression.\n\nDo not allow `ROADMAP.md` to grow endlessly.\n\nOn every pass:\n\n* merge duplicate items\n* remove stale wording\n* shorten old sections\n* archive outdated ideas\n* preserve only useful strategic memory\n* keep Now small\n* keep Next realistic\n* keep Later intentional\n* keep Discovery honest\n\nIf a section becomes too large, summarize it.\n\nIf an item has no next action, confidence, trigger, or evidence, it probably does not belong in Now or Next.\n\n## Non-Technical Navigation Rules\n\nThe roadmap must be readable by non-technical users.\n\nUse plain language first.\n\nTechnical details may appear only when they clarify direction, risk, or ownership.\n\nPrefer:\n\n```text\n\"Create one canonical patch pathway for agents.\"\n```\n\nOver:\n\n```text\n\"Refactor RPC mutation handler abstractions across adapter modules.\"\n```\n\nThe implementation details can live in issues, commits, or engineering docs.\n\nThe roadmap should explain why the work matters.\n\n## Agent Governance Rules\n\nAgents using this roadmap must optimize for coherence before completion.\n\nAgents should:\n\n* reuse existing architecture\n* extend canonical systems\n* preserve naming patterns\n* avoid parallel workflows\n* centralize diagnostics\n* centralize mutation paths\n* reduce hidden behavior\n* improve inspectability\n* document decisions\n* archive stale ideas\n\nAgents must avoid:\n\n* inventing new architecture without justification\n* creating duplicate utility layers\n* creating local orchestration systems\n* scattering state ownership\n* hiding behavior behind convenience wrappers\n* adding speculative roadmap items as commitments\n* turning Discovery into Now without evidence\n* treating the roadmap as a task dump\n\n## Code Soup Detection Heuristics\n\nFlag code soup risk when any of the following appear:\n\n* multiple files doing the same job\n* multiple commands for the same workflow\n* multiple sources of truth\n* repeated helper functions\n* unclear ownership of state\n* new abstractions with no clear gravity well\n* duplicated configuration\n* inconsistent naming\n* hidden side effects\n* new scripts bypassing canonical runtime behavior\n* agents repeatedly patching the same area differently\n* debugging requires jumping across unrelated systems\n\nWhen detected, recommend convergence.\n\nConvergence may mean:\n\n* merge duplicate paths\n* document the canonical path\n* delete dead paths\n* create a single command surface\n* route behavior through the existing runtime\n* centralize diagnostics\n* collapse abstractions\n* rename for clarity\n* archive roadmap work that causes fragmentation\n\n## Update Style\n\nWhen editing `ROADMAP.md`:\n\n* be concise\n* be specific\n* preserve useful history\n* remove noise\n* use stable headings\n* avoid hype\n* avoid generic PM language\n* avoid massive bullet lists unless necessary\n* prefer fewer, stronger items\n* make uncertainty visible\n* make next actions concrete\n\nThe tone should feel like a calm senior product strategist, staff engineer, and technical program manager reviewed the system together.\n\n## Required Final Assistant Response\n\nAfter updating or drafting `ROADMAP.md`, respond with this summary:\n\n```markdown\n## Roadmap Checkpoint Updated\n\n**Health:** \u003Cstatus>\n\n**Center of Gravity:**  \n\u003Cone-sentence summary>\n\n**Moved:**  \n- \u003Citems moved between sections, or \"None\">\n\n**Added:**  \n- \u003Cnew items, or \"None\">\n\n**Updated:**  \n- \u003Cupdated items, or \"None\">\n\n**Archived:**  \n- \u003Carchived items, or \"None\">\n\n**Code Soup Risk:** Low \u002F Medium \u002F High  \n\u003Cbrief reason>\n\n**Recommended Next Move:**  \n\u003Cone concrete action>\n```\n\nDo not include the full roadmap in the final response unless the user asks for it.\n\nThe final response should summarize the checkpoint, not duplicate the file.\n\n## Failure Modes\n\nThis skill fails if it:\n\n* creates a generic roadmap\n* produces a giant backlog\n* ignores center of gravity\n* skips the code soup audit\n* invents project state\n* hides uncertainty\n* promotes vague ideas into Now\n* deletes strategic memory without reason\n* decentralizes authority without mitigation\n* makes the roadmap harder to read\n* treats implementation activity as automatic strategic progress\n\n## Success Definition\n\nThis skill succeeds when `ROADMAP.md` becomes the project's living steering checkpoint.\n\nA successful roadmap should:\n\n* preserve strategic memory\n* clarify current direction\n* prevent code soup\n* expose maintenance gravity\n* guide agents toward canonical paths\n* make long-term development navigable\n* help non-technical users understand what matters\n* help technical users understand where changes belong\n* reduce cognitive burden over time\n\nThe highest goal is not more planning.\n\nThe highest goal is sustained coherence under long-horizon, agent-assisted development.\n\n## DietCode Plugin Integration\n\nWhen the DietCode plugin is active:\n\n1. Call `roadmap(action='guide')` to learn phase, health, `_roadmap_operator_hints`, and `agent_next_call`.\n2. Call `roadmap(action='cockpit')` or `\u002Froadmap cockpit` for a one-screen operator summary.\n3. Call `roadmap(action='checkpoint', context=…)` before editing `ROADMAP.md` — returns evidence, `code_soup_pre_audit`, and the 16-step algorithm.\n4. Call `roadmap(action='template')` when bootstrapping the first `ROADMAP.md` — returns `project_steering_digest` and `bootstrap_autofill_preview`.\n5. When bootstrap template phrases remain, call `roadmap(action='apply_bootstrap_fill')` to preview evidence-backed replacements; pass `context='write'` to apply, then `roadmap(action='validate')`.\n6. Use `project_steering_digest` and `bootstrap_fill_plan.tasks[].suggested_replacement` for per-project fill — each task maps template text to README\u002Fgit\u002Ffingerprint evidence.\n7. After editing, call `roadmap(action='validate')` to confirm schema compliance before finishing.\n8. Call `roadmap(action='doctor')` to install the skill and run production health checks.\n9. Call `roadmap(action='evidence')` for a read-only evidence bundle with `project_fingerprint`.\n10. Call `roadmap(action='status')` to parse the current roadmap without mutating it.\n11. Call `roadmap(action='explain_gate')` or `\u002Froadmap explain-gate` when schema or freshness gates block progress (kernel explain-gate analogue).\n12. Call `joyzoning(action='roadmap')` for a native cockpit brief inside governed sessions.\n\nWorkspace state persists at `.dietcode\u002Froadmap-state.json` after each `validate` pass.\n\nThe bundled default skill (`auto-rolling-roadmap`) ships with the extension and is registered automatically when `dietcode.roadmap.auto_install_skills` is enabled (default: true). It is **not** copied into workspaces.\n\n**Throughput model (industry-standard progressive disclosure):**\n- **Metadata** — listed in the SKILLS prompt section for optional user skills only; bundled roadmap is omitted when ROADMAP_STEERING is active (no duplicate surface).\n- **Instructions** — loaded on-demand via `use_skill`; bundled roadmap returns a lean execution digest (~30 lines), not the full 962-line reference.\n- **Reference** — full SKILL.md remains bundled for deep checkpoint algorithm detail when needed.\n\n**Execution authority:** Continue the user's I\u002FO loop. Roadmap governance clears at `attempt_completion` — no mid-task validate\u002Fdoctor unless completion is blocked.\n\nOperator smoke: `python scripts\u002Froadmap_smoke.py` · `python scripts\u002Froadmap_operator_smoke.py`\n","# DietCode Skill: Auto-Rolling Roadmap Checkpoint System\n\n## Skill Purpose\n\nMaintain an evolving `ROADMAP.md` file that acts as the project's living product, architecture, and long-horizon development checkpoint.\n\nThis skill exists to prevent code soup, preserve project coherence, and keep long-running development aligned around a clear center of gravity.\n\nThe roadmap is not a wishlist.\n\nThe roadmap is not a backlog.\n\nThe roadmap is not a promise.\n\nThe roadmap is the project's steering surface.\n\nIt should help humans and agents answer:\n\n* What is this project becoming?\n* What matters now?\n* What should happen next?\n* What should remain deferred?\n* What changed recently?\n* What risks are accumulating?\n* Is the system becoming more coherent or more fragmented?\n\n## Primary Output\n\nThis skill must create or update:\n\n```text\nROADMAP.md\n```\n\n`ROADMAP.md` is the canonical checkpoint file for long-horizon project steering.\n\nIf the file does not exist, create it.\n\nIf the file exists, evolve it carefully.\n\nDo not replace useful history with generic text.\n\nDo not append endlessly.\n\nDo not turn the roadmap into a backlog dump.\n\nEach update should make the roadmap more useful, more current, and easier to navigate.\n\n## Prime Directive\n\nEvery roadmap pass must answer:\n\n```text\nDid the latest work strengthen or weaken the project's center of gravity?\n```\n\nIf the answer is unclear, the skill must investigate before adding or promoting roadmap items.\n\nA roadmap item that does not connect to the project's center of gravity should remain in Discovery, be reframed, or be archived.\n\n## Definition: Center of Gravity\n\nThe project's center of gravity is the smallest set of concepts, workflows, abstractions, and operational paths that explain how the system fundamentally works.\n\nA strong center of gravity means:\n\n* there is one obvious way to perform core operations\n* agents can predict where changes belong\n* contributors can understand the system without reading everything\n* new features attach to existing architecture instead of creating parallel systems\n* operational behavior is inspectable\n* maintenance burden stays bounded\n* long-term development feels navigable\n\nA weak center of gravity means:\n\n* duplicate abstractions appear\n* multiple workflows compete\n* state ownership becomes unclear\n* debugging paths fragment\n* agents invent local patterns\n* hidden orchestration grows\n* roadmap direction becomes disconnected from implementation reality\n\n## Core Philosophy\n\nCentralization does not mean giant files, giant classes, or rigid architecture.\n\nCentralization means clear authority.\n\nModules may be distributed.\n\nAuthority should not be.\n\nThe roadmap should continuously protect:\n\n* canonical workflows\n* canonical mutation paths\n* canonical runtime inspection\n* canonical state ownership\n* canonical terminology\n* canonical architectural intent\n\nThe system should feel extensible without becoming incoherent.\n\n## Required Inputs\n\nWhen invoked, inspect available project context before editing `ROADMAP.md`.\n\nUse any available sources such as:\n\n* existing `ROADMAP.md`\n* README files\n* architecture docs\n* recent commits\n* changed files\n* tests\n* scripts\n* issue notes\n* implementation summaries\n* failing commands\n* current user request\n* agent output logs\n* TODOs or FIXME markers\n* package\u002Fconfig changes\n\nDo not rely only on the user's latest request if project artifacts are available.\n\nDo not invent project state.\n\nIf evidence is missing, mark uncertainty explicitly.\n\n## Per-Project Identity (DietCode native integration)\n\nEvery `roadmap` tool response is scoped to the **Hermes project workspace**, not\nthe DietCode plugin install tree. Read these fields before editing:\n\n| Field | Use |\n| --- | --- |\n| `project_identity_line` | One-line header: project brief · stack · verify command |\n| `project_steering_digest` | Entity card: CI, quality tools, governance, verify, bootstrap status |\n| `project_fingerprint` | Raw signals in checkpoint `evidence` (README title, archetype, Makefile targets, …) |\n| `bootstrap_fill_plan` | When template phrases remain — use `tasks[].suggested_replacement` |\n\nPrefer evidence-backed replacements over generic text. When placeholders remain,\ncall `roadmap(action='apply_bootstrap_fill')` before manual edits.\n\nOperator reference: `docs\u002Froadmap.md` in the DietCode plugin tree.\n\nKey doc sections for agents:\n\n- **12-section schema** — required headings and enumerated health\u002Fsoup values\n- **Example payloads** — shape of `guide`, `checkpoint`, and `validate` responses\n- **Bootstrap fill** — `bootstrap_fill_plan.tasks[].suggested_replacement`\n- **Write guard** — ROADMAP.md only at `{workspace}\u002FROADMAP.md`\n- **Anti-patterns** — backlog dumping, skipping §9 audit, plugin-tree writes\n\n## Required Update Algorithm\n\nFollow this sequence on every roadmap pass:\n\n1. Read the existing `ROADMAP.md`, if present.\n2. Identify the current stated center of gravity.\n3. Inspect recent project changes and available evidence.\n4. Determine whether the project is coherent, accelerating, drifting, fragmenting, blocked, or overloaded.\n5. Compare current work against the existing roadmap.\n6. Preserve valid strategic intent.\n7. Remove duplicate or stale roadmap entries.\n8. Archive items that no longer deserve active attention.\n9. Promote items only when evidence supports promotion.\n10. Demote items when uncertainty, risk, or entropy increases.\n11. Add new items only when they connect to the center of gravity.\n12. Rewrite technical implementation details into clear product and architecture language.\n13. Run a centralization and code soup audit.\n14. Update the recent checkpoint.\n15. Add decision log entries for meaningful direction changes.\n16. Return a concise summary of what changed.\n\n## ROADMAP.md Required Structure\n\nThe file should use this structure.\n\n```markdown\n# ROADMAP.md\n\n## 1. Project Center of Gravity\n\n## 2. Roadmap Health\n\n## 3. Strategic Narrative\n\n## 4. Now\n\n## 5. Next\n\n## 6. Later\n\n## 7. Discovery\n\n## 8. Maintenance Gravity\n\n## 9. Centralization & Code Soup Audit\n\n## 10. Decision Log\n\n## 11. Recent Checkpoint\n\n## 12. Archive\n```\n\nDo not rename these top-level sections unless the user explicitly requests a different schema.\n\nThe consistency of the schema is part of the skill's value.\n\n## Section Contract\n\n### 1. Project Center of Gravity\n\nDescribe what the project fundamentally is.\n\nThis section should be understandable to a non-technical stakeholder while still useful to an engineer or agent.\n\nInclude:\n\n```markdown\n## 1. Project Center of Gravity\n\n**Core Purpose:**  \n\u003Cplain-language purpose>\n\n**Primary Users \u002F Operators:**  \n\u003Cwho uses or operates this system>\n\n**Canonical Architecture:**  \n\u003Cshort description of the main architectural shape>\n\n**Canonical Workflows:**  \n\u003Cthe main flows agents and humans should preserve>\n\n**Primary Runtime \u002F Operational Center:**  \n\u003Cwhere operational truth lives>\n\n**What This Project Must Not Become:**  \n\u003Canti-goals that protect coherence>\n```\n\nThe \"must not become\" field is mandatory.\n\nIt prevents drift.\n\nExamples:\n\n* must not become a scattered collection of scripts\n* must not become another generic AI IDE\n* must not create multiple competing orchestration systems\n* must not hide mutations behind invisible automation\n* must not trade inspectability for convenience\n\n### 2. Roadmap Health\n\nGive the current health state.\n\nAllowed statuses:\n\n```text\nCoherent\nAccelerating\nDrifting\nFragmenting\nBlocked\nOverloaded\nRecovering\n```\n\nUse exactly one primary status.\n\nFormat:\n\n```markdown\n## 2. Roadmap Health\n\n**Status:** Coherent \u002F Accelerating \u002F Drifting \u002F Fragmenting \u002F Blocked \u002F Overloaded \u002F Recovering\n\n**Summary:**  \n\u003Cbrief explanation>\n\n**Why This Status:**  \n- \u003Cevidence>\n- \u003Cevidence>\n- \u003Cevidence>\n\n**Primary Risk:**  \n\u003Csingle biggest risk>\n\n**Primary Opportunity:**  \n\u003Csingle biggest opportunity>\n```\n\n### 3. Strategic Narrative\n\nExplain what the project is becoming.\n\nThis should be short, clear, and directional.\n\nAvoid vague hype.\n\nFormat:\n\n```markdown\n## 3. Strategic Narrative\n\n\u003C2-5 paragraphs explaining the current product \u002F architecture direction>\n\nThe narrative should clarify:\n- what changed recently\n- what the system is converging toward\n- what should be protected\n- what should be avoided\n```\n\n### 4. Now\n\nActive, committed, or already-in-motion work.\n\nOnly include items that are ready for execution or currently underway.\n\nDo not include vague ideas.\n\nEach item must use this format:\n\n```markdown\n### N. \u003CItem Title>\n\n**Goal:**  \n\u003Cwhat this work accomplishes>\n\n**Why It Matters:**  \n\u003Cproduct, architecture, user, or maintenance value>\n\n**Current State:**  \n\u003Cknown state from evidence>\n\n**Next Concrete Action:**  \n\u003Cone specific next action>\n\n**Success Signal:**  \n\u003Chow we know this worked>\n\n**Gravity Impact:** Strengthens \u002F Neutral \u002F Weakens \u002F Unknown  \n**Centralization Effect:** Centralizes \u002F No Change \u002F Decentralizes  \n**Entropy Risk:** Low \u002F Medium \u002F High  \n\n**Risk \u002F Mitigation:**  \n\u003Crisk and mitigation>\n```\n\nRules for Now:\n\n* keep it small\n* prefer 1-5 items\n* each item must be actionable\n* each item must have evidence\n* each item must connect to the center of gravity\n\nIf there are more than 5 Now items, the roadmap is overloaded.\n\n### 5. Next\n\nLikely upcoming work.\n\nThese are prepared but not fully committed.\n\nFormat:\n\n```markdown\n### N. \u003CItem Title>\n\n**Opportunity:**  \n\u003Cwhat could be unlocked>\n\n**Why Soon:**  \n\u003Cwhy this may matter after Now>\n\n**Dependency:**  \n\u003Cwhat must happen first>\n\n**First Validation Step:**  \n\u003Csmallest way to test whether this deserves promotion>\n\n**Confidence:** High \u002F Medium \u002F Low  \n\n**Gravity Impact:** Strengthens \u002F Neutral \u002F Weakens \u002F Unknown  \n**Centralization Effect:** Centralizes \u002F No Change \u002F Decentralizes  \n**Entropy Risk:** Low \u002F Medium \u002F High\n```\n\nRules for Next:\n\n* do not pretend Next is committed\n* include validation steps\n* use confidence labels\n* avoid implementation rabbit holes\n\n### 6. Later\n\nLong-horizon possibilities.\n\nThese are strategic directions, not promises.\n\nFormat:\n\n```markdown\n### N. \u003CDirection Title>\n\n**Direction:**  \n\u003Cwhat this could become>\n\n**Potential Upside:**  \n\u003Cwhy it may matter>\n\n**Why Not Now:**  \n\u003Cwhy it remains deferred>\n\n**Promotion Trigger:**  \n\u003Cwhat evidence would move this into Next>\n\n**Entropy Risk:** Low \u002F Medium \u002F High\n```\n\nRules for Later:\n\n* keep speculative work out of Now\n* preserve interesting possibilities without letting them hijack execution\n* include promotion triggers so Later does not become a junk drawer\n\n### 7. Discovery\n\nQuestions, experiments, unknowns, and ambiguous ideas.\n\nDiscovery exists to protect the roadmap from false certainty.\n\nFormat:\n\n```markdown\n### N. \u003CQuestion or Investigation>\n\n**Question:**  \n\u003Cwhat needs to be understood>\n\n**Evidence Needed:**  \n\u003Cwhat would clarify this>\n\n**Possible Outcomes:**  \n- \u003Coutcome>\n- \u003Coutcome>\n- \u003Coutcome>\n\n**Decision Needed:**  \n\u003Cwhat decision this discovery should eventually support>\n```\n\nRules for Discovery:\n\n* vague ideas belong here, not Now\n* technical uncertainty belongs here\n* product uncertainty belongs here\n* architectural ambiguity belongs here\n\n### 8. Maintenance Gravity\n\nTrack places where complexity, confusion, or fragility is accumulating.\n\nFormat:\n\n```markdown\n## 8. Maintenance Gravity\n\n### Hotspots\n\n| Area | Symptom | Risk | Recommended Action |\n|---|---|---|---|\n| \u003Carea> | \u003Csymptom> | Low\u002FMedium\u002FHigh | \u003Caction> |\n\n### Repeated Friction\n\n- \u003Cbug, workflow issue, agent confusion, or maintenance pattern>\n\n### Documentation Gaps\n\n- \u003Cmissing or stale docs>\n\n### Agent Confusion Points\n\n- \u003Cplaces agents are likely to patch incorrectly>\n```\n\nMaintenance gravity is not shame.\n\nIt is early warning.\n\nThe purpose is to detect entropy before it becomes structural collapse.\n\n### 9. Centralization & Code Soup Audit\n\nEvaluate whether the project is becoming more or less coherent.\n\nUse this checklist:\n\n```markdown\n## 9. Centralization & Code Soup Audit\n\n**Overall Code Soup Risk:** Low \u002F Medium \u002F High\n\n### Canonical Path Integrity\n\n- Are there multiple ways to perform the same operation?\n- Are there duplicate APIs, scripts, commands, or workflows?\n- Is the correct path obvious?\n\n**Assessment:**  \n\u003Canswer>\n\n### Authority Boundaries\n\n- Where does runtime authority live?\n- Where does state authority live?\n- Where does mutation authority live?\n- Where does diagnostic authority live?\n\n**Assessment:**  \n\u003Canswer>\n\n### Structural Drift\n\n- Did recent work introduce isolated patterns?\n- Did it create one-off abstractions?\n- Did it bypass existing architecture?\n\n**Assessment:**  \n\u003Canswer>\n\n### Agent Coherence\n\n- Would a future agent know where to patch?\n- Would a future agent reuse the canonical path?\n- Did recent work make agent behavior safer or more chaotic?\n\n**Assessment:**  \n\u003Canswer>\n\n### Centralization Recommendation\n\n\u003Cone recommendation to strengthen project gravity>\n```\n\nThis section is mandatory.\n\nIf the roadmap does not audit code soup risk, the skill failed.\n\n### 10. Decision Log\n\nRecord meaningful product, architecture, and workflow decisions.\n\nFormat:\n\n```markdown\n## 10. Decision Log\n\n### YYYY-MM-DD — \u003CDecision Title>\n\n**Decision:**  \n\u003Cwhat was decided>\n\n**Reason:**  \n\u003Cwhy>\n\n**Impact:**  \n\u003Cwhat changes because of this>\n\n**Follow-up:**  \n\u003Cwhat should happen next>\n```\n\nRules:\n\n* add entries only for meaningful decisions\n* do not log every tiny change\n* do not rewrite history casually\n* preserve old decisions unless they are clearly obsolete\n* if a decision is reversed, add a new entry instead of deleting the old one\n\n### 11. Recent Checkpoint\n\nSummarize the latest roadmap update.\n\nFormat:\n\n```markdown\n## 11. Recent Checkpoint\n\n**Date:** YYYY-MM-DD\n\n**Checkpoint Summary:**  \n\u003Cwhat changed this pass>\n\n**Moved:**  \n- \u003Citem moved from one section to another>\n\n**Added:**  \n- \u003Cnew item>\n\n**Updated:**  \n- \u003Cexisting item updated>\n\n**Archived:**  \n- \u003Carchived item>\n\n**Code Soup Risk:** Low \u002F Medium \u002F High  \n\u003Cbrief reason>\n\n**Recommended Next Move:**  \n\u003Cone concrete action>\n```\n\nOnly one Recent Checkpoint should exist.\n\nWhen updating, replace the previous Recent Checkpoint with the latest one.\n\nOlder meaningful changes should be preserved in the Decision Log or Archive.\n\n### 12. Archive\n\nThe archive stores removed, deferred, killed, or superseded roadmap items.\n\nFormat:\n\n```markdown\n## 12. Archive\n\n### \u003CArchived Item>\n\n**Archived Date:** YYYY-MM-DD  \n**Reason:**  \n\u003Cwhy it was removed from active attention>\n\n**Restore Condition:**  \n\u003Cwhat would justify bringing it back>\n```\n\nRules:\n\n* archive instead of deleting when strategic memory matters\n* delete only if the item was duplicate, noise, or clearly useless\n* archive stale speculative work\n* archive work that weakens center of gravity without sufficient upside\n\n## Roadmap Item Scoring\n\nEvery Now, Next, and Later item must include gravity metadata.\n\nUse these exact labels.\n\n### Gravity Impact\n\n```text\nStrengthens\nNeutral\nWeakens\nUnknown\n```\n\nMeaning:\n\n* Strengthens: reinforces the project center of gravity\n* Neutral: does not meaningfully affect coherence\n* Weakens: fragments authority, duplicates systems, or increases drift\n* Unknown: not enough evidence yet\n\n### Centralization Effect\n\n```text\nCentralizes\nNo Change\nDecentralizes\n```\n\nMeaning:\n\n* Centralizes: moves authority, diagnostics, workflow, or state closer to canonical surfaces\n* No Change: does not affect centralization\n* Decentralizes: creates new authority surfaces, parallel paths, or hidden behavior\n\n### Entropy Risk\n\n```text\nLow\nMedium\nHigh\n```\n\nMeaning:\n\n* Low: unlikely to create maintenance burden\n* Medium: manageable but requires attention\n* High: likely to create fragmentation, confusion, or long-term maintenance cost\n\n## Promotion and Demotion Rules\n\n### Promote Discovery to Next only when:\n\n* the question is clear\n* evidence exists\n* the opportunity connects to the center of gravity\n* the first validation step is known\n* entropy risk is understood\n\n### Promote Next to Now only when:\n\n* dependencies are resolved\n* the next action is concrete\n* success signal is clear\n* the work is aligned with current strategy\n* the roadmap has capacity\n\n### Demote Now to Next when:\n\n* work is blocked\n* confidence drops\n* dependencies are missing\n* implementation risk grows\n* it no longer fits the current center of gravity\n\n### Move anything to Archive when:\n\n* it is stale\n* it duplicates another item\n* it weakens project gravity\n* it belongs to an older strategy\n* it is not worth preserving in active attention\n\n## Anti-Bloat Rules\n\nThe roadmap must evolve by compression.\n\nDo not allow `ROADMAP.md` to grow endlessly.\n\nOn every pass:\n\n* merge duplicate items\n* remove stale wording\n* shorten old sections\n* archive outdated ideas\n* preserve only useful strategic memory\n* keep Now small\n* keep Next realistic\n* keep Later intentional\n* keep Discovery honest\n\nIf a section becomes too large, summarize it.\n\nIf an item has no next action, confidence, trigger, or evidence, it probably does not belong in Now or Next.\n\n## Non-Technical Navigation Rules\n\nThe roadmap must be readable by non-technical users.\n\nUse plain language first.\n\nTechnical details may appear only when they clarify direction, risk, or ownership.\n\nPrefer:\n\n```text\n\"Create one canonical patch pathway for agents.\"\n```\n\nOver:\n\n```text\n\"Refactor RPC mutation handler abstractions across adapter modules.\"\n```\n\nThe implementation details can live in issues, commits, or engineering docs.\n\nThe roadmap should explain why the work matters.\n\n## Agent Governance Rules\n\nAgents using this roadmap must optimize for coherence before completion.\n\nAgents should:\n\n* reuse existing architecture\n* extend canonical systems\n* preserve naming patterns\n* avoid parallel workflows\n* centralize diagnostics\n* centralize mutation paths\n* reduce hidden behavior\n* improve inspectability\n* document decisions\n* archive stale ideas\n\nAgents must avoid:\n\n* inventing new architecture without justification\n* creating duplicate utility layers\n* creating local orchestration systems\n* scattering state ownership\n* hiding behavior behind convenience wrappers\n* adding speculative roadmap items as commitments\n* turning Discovery into Now without evidence\n* treating the roadmap as a task dump\n\n## Code Soup Detection Heuristics\n\nFlag code soup risk when any of the following appear:\n\n* multiple files doing the same job\n* multiple commands for the same workflow\n* multiple sources of truth\n* repeated helper functions\n* unclear ownership of state\n* new abstractions with no clear gravity well\n* duplicated configuration\n* inconsistent naming\n* hidden side effects\n* new scripts bypassing canonical runtime behavior\n* agents repeatedly patching the same area differently\n* debugging requires jumping across unrelated systems\n\nWhen detected, recommend convergence.\n\nConvergence may mean:\n\n* merge duplicate paths\n* document the canonical path\n* delete dead paths\n* create a single command surface\n* route behavior through the existing runtime\n* centralize diagnostics\n* collapse abstractions\n* rename for clarity\n* archive roadmap work that causes fragmentation\n\n## Update Style\n\nWhen editing `ROADMAP.md`:\n\n* be concise\n* be specific\n* preserve useful history\n* remove noise\n* use stable headings\n* avoid hype\n* avoid generic PM language\n* avoid massive bullet lists unless necessary\n* prefer fewer, stronger items\n* make uncertainty visible\n* make next actions concrete\n\nThe tone should feel like a calm senior product strategist, staff engineer, and technical program manager reviewed the system together.\n\n## Required Final Assistant Response\n\nAfter updating or drafting `ROADMAP.md`, respond with this summary:\n\n```markdown\n## Roadmap Checkpoint Updated\n\n**Health:** \u003Cstatus>\n\n**Center of Gravity:**  \n\u003Cone-sentence summary>\n\n**Moved:**  \n- \u003Citems moved between sections, or \"None\">\n\n**Added:**  \n- \u003Cnew items, or \"None\">\n\n**Updated:**  \n- \u003Cupdated items, or \"None\">\n\n**Archived:**  \n- \u003Carchived items, or \"None\">\n\n**Code Soup Risk:** Low \u002F Medium \u002F High  \n\u003Cbrief reason>\n\n**Recommended Next Move:**  \n\u003Cone concrete action>\n```\n\nDo not include the full roadmap in the final response unless the user asks for it.\n\nThe final response should summarize the checkpoint, not duplicate the file.\n\n## Failure Modes\n\nThis skill fails if it:\n\n* creates a generic roadmap\n* produces a giant backlog\n* ignores center of gravity\n* skips the code soup audit\n* invents project state\n* hides uncertainty\n* promotes vague ideas into Now\n* deletes strategic memory without reason\n* decentralizes authority without mitigation\n* makes the roadmap harder to read\n* treats implementation activity as automatic strategic progress\n\n## Success Definition\n\nThis skill succeeds when `ROADMAP.md` becomes the project's living steering checkpoint.\n\nA successful roadmap should:\n\n* preserve strategic memory\n* clarify current direction\n* prevent code soup\n* expose maintenance gravity\n* guide agents toward canonical paths\n* make long-term development navigable\n* help non-technical users understand what matters\n* help technical users understand where changes belong\n* reduce cognitive burden over time\n\nThe highest goal is not more planning.\n\nThe highest goal is sustained coherence under long-horizon, agent-assisted development.\n\n## DietCode Plugin Integration\n\nWhen the DietCode plugin is active:\n\n1. Call `roadmap(action='guide')` to learn phase, health, `_roadmap_operator_hints`, and `agent_next_call`.\n2. Call `roadmap(action='cockpit')` or `\u002Froadmap cockpit` for a one-screen operator summary.\n3. Call `roadmap(action='checkpoint', context=…)` before editing `ROADMAP.md` — returns evidence, `code_soup_pre_audit`, and the 16-step algorithm.\n4. Call `roadmap(action='template')` when bootstrapping the first `ROADMAP.md` — returns `project_steering_digest` and `bootstrap_autofill_preview`.\n5. When bootstrap template phrases remain, call `roadmap(action='apply_bootstrap_fill')` to preview evidence-backed replacements; pass `context='write'` to apply, then `roadmap(action='validate')`.\n6. Use `project_steering_digest` and `bootstrap_fill_plan.tasks[].suggested_replacement` for per-project fill — each task maps template text to README\u002Fgit\u002Ffingerprint evidence.\n7. After editing, call `roadmap(action='validate')` to confirm schema compliance before finishing.\n8. Call `roadmap(action='doctor')` to install the skill and run production health checks.\n9. Call `roadmap(action='evidence')` for a read-only evidence bundle with `project_fingerprint`.\n10. Call `roadmap(action='status')` to parse the current roadmap without mutating it.\n11. Call `roadmap(action='explain_gate')` or `\u002Froadmap explain-gate` when schema or freshness gates block progress (kernel explain-gate analogue).\n12. Call `joyzoning(action='roadmap')` for a native cockpit brief inside governed sessions.\n\nWorkspace state persists at `.dietcode\u002Froadmap-state.json` after each `validate` pass.\n\nThe bundled default skill (`auto-rolling-roadmap`) ships with the extension and is registered automatically when `dietcode.roadmap.auto_install_skills` is enabled (default: true). It is **not** copied into workspaces.\n\n**Throughput model (industry-standard progressive disclosure):**\n- **Metadata** — listed in the SKILLS prompt section for optional user skills only; bundled roadmap is omitted when ROADMAP_STEERING is active (no duplicate surface).\n- **Instructions** — loaded on-demand via `use_skill`; bundled roadmap returns a lean execution digest (~30 lines), not the full 962-line reference.\n- **Reference** — full SKILL.md remains bundled for deep checkpoint algorithm detail when needed.\n\n**Execution authority:** Continue the user's I\u002FO loop. Roadmap governance clears at `attempt_completion` — no mid-task validate\u002Fdoctor unless completion is blocked.\n\nOperator smoke: `python scripts\u002Froadmap_smoke.py` · `python scripts\u002Froadmap_operator_smoke.py`\n",{"name":5,"description":7},"main","4930239eb4257d5060fa17753ceae9a81efd4e00","optional-skills\u002Fdietcode\u002Fauto-rolling-roadmap",0,"https:\u002F\u002Fgithub.com\u002FCardSorting\u002FHEAV3NS","https:\u002F\u002Fgithub.com\u002FCardSorting",";)","2026-09-30 03:14:15",{"source":8,"skillName":5,"commands":33,"prompt":43,"zipAvailable":44},[34,37,40],{"tool":35,"command":36},"npx","npx skills add CardSorting\u002FHEAV3NS --skill auto-rolling-roadmap",{"tool":38,"command":39},"bunx","bunx skills add CardSorting\u002FHEAV3NS --skill auto-rolling-roadmap",{"tool":41,"command":42},"pnpm","pnpm dlx skills add CardSorting\u002FHEAV3NS --skill auto-rolling-roadmap","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002FCardSorting\u002FHEAV3NS\u002Fblob\u002Fmain\u002Foptional-skills\u002Fdietcode\u002Fauto-rolling-roadmap\u002FSKILL.md",true,[],[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},"01M7CX14MK927TPNW3YEY7MY6J","starter-skill","Foundational skill template demonstrating procedural steps and safety guardrails.","https:\u002F\u002Fgithub.com\u002FCardSorting\u002FHEAV3NS\u002Fblob\u002Fmain\u002Fskills\u002Ftemplates\u002Fstarter-skill\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},"01M7CX124ARA06RX0ZZ1XN7FYT","security-sentinel","Security sentinel that analyzes ASTs for injection risks, verifies file path sandboxing, and prevents command tampering.","https:\u002F\u002Fgithub.com\u002FCardSorting\u002FHEAV3NS\u002Fblob\u002Fmain\u002Fskills\u002Fsecurity-sentinel\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},"01M7CX0YVV7A9GYG3S0GGC8P2A","performance-auditor","A high-throughput performance auditor that verifies zero-GC boundaries, memory slab invariants, and hot-path execution latencies.","https:\u002F\u002Fgithub.com\u002FCardSorting\u002FHEAV3NS\u002Fblob\u002Fmain\u002Fskills\u002Fperformance-auditor\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},"01M7CX0VTAFVFEVWRYRA9WH1HX","code-refactor","Clean architecture refactoring specialist enforcing SOLID principles, strict TypeScript interfaces, and zero barrel imports.","https:\u002F\u002Fgithub.com\u002FCardSorting\u002FHEAV3NS\u002Fblob\u002Fmain\u002Fskills\u002Fcode-refactor\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},"01M7CX0RXQA0AS4MBEW321NQCY","adversarial-auditor","Adversarial auditor that red-teams technical plans, asserts fail-closed factual provenance, decomposes cognitive token spend, and prevents premature completion.","https:\u002F\u002Fgithub.com\u002FCardSorting\u002FHEAV3NS\u002Fblob\u002Fmain\u002Fskills\u002Fadversarial-auditor\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},"01M7CX0NH6XSMSADS7Z58Q0P3J","golden-cartridge-protocol","Optimize for verified engineering value with finite context, execution, mutation, and review capacity.","https:\u002F\u002Fgithub.com\u002FCardSorting\u002FHEAV3NS\u002Fblob\u002Fmain\u002Foptional-skills\u002Fdietcode\u002Fgolden-cartridge-protocol\u002FSKILL.md",[],[]]