agentsclimarketplace

Project maintainer

Skill Pratiyush/llm-wiki/.claude/skills/project-maintainer

Maintain an llmwiki-style open-source project across the full framework pipeline. Use when the user says "maintain the project", "check my tasks", "update progress", "run phase gate", "do a monthly verify", "lint my wiki", "check stale entries", or invokes any form of ongoing project-keeping chore. Reads _progress.md, tasks.md, docs/roadmap.md, and .github issues to figure out what's next and what's drifted.From its SKILL.md

Install
npx -y skills add Pratiyush/llm-wiki --skill project-maintainer

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

SKILL.md

5.7 KB, as published. Nobody here has run it

project-maintainer

What this skill does

The Maintainer owns the full Open Source Framework v4.1 pipeline on a project. Given the current repo state, it:

  1. Audits phase gates — are we in the right phase? Are the deliverables for the current phase complete?
  2. Reconciles task tracking — keeps tasks.md, _progress.md, and GitHub issues in sync (the framework's "dual-tracking" rule).
  3. Runs stale checks — flags content, links, and claims older than the stale threshold.
  4. Triggers CI — re-runs the privacy grep, performance budget check, and link checker.
  5. Reports next action — tells the user exactly what to do next and why.

It works for any project that follows the framework — not just llmwiki.

When to invoke

  • "Maintain the project" / "maintenance check" / "do a health check"
  • "Update progress" / "reconcile tasks" / "sync tasks"
  • "Check phase gate" / "can I move to the next phase"
  • "Lint the wiki" / "check for stale entries"
  • "Monthly verify" / "quarterly review"
  • Any time the user is unsure what to work on next

Framework phase awareness

PhaseThe Maintainer's job in this phase
0 CaptureVerify idea-brief.md exists and names target users + the 10x mechanism
1 ValidateVerify scorecard /25 is in _progress.md and ≥20
1.25 ResearchVerify .temp/ has at least 10 cloned reference repos and docs/research.md exists with a gap matrix
1.5 SteeringVerify _progress.md has "Key Decisions" table filled in
1.75 Agent SurveyFor agent-native tools — verify adapter compatibility matrix exists
2 BrandVerify README.md, LICENSE, and badges are in place
3 StructureVerify folder layout matches docs/architecture.md
4 ContentTrack which M items in docs/roadmap.md are ✅ vs [ ]
5 ContributionVerify CONTRIBUTING.md, PR template, issue templates, CI workflow
5.25 Adapter FlowFor agent-native tools — verify adapter contract doc + at least one community contribution example
5.5 Pre-Launch QARun the full QA checklist from docs/framework.md
6 LaunchVerify git tag + GitHub Release + social posts drafted
6.5 Self-DemoVerify GitHub Pages workflow triggered and live URL returns 200
7 GrowTrack stars / forks / downloads week-over-week
7.5 Living KnowledgeVerify public wiki is updating on release
8 MaintainRun monthly verification, merge PRs, update stale entries

Workflow

  1. Locate the project root. Look for _progress.md in the current dir or its parents. If missing, tell the user this skill needs a project that follows the framework.

  2. Read the state files:

    • _progress.md — current phase + phase status table
    • tasks.md — Kiro-style tasks with [ ]/[/]/[x]/[-] markers
    • docs/roadmap.md — if present, the master MoSCoW table
    • CHANGELOG.md — latest release
  3. Audit each phase gate up to the current phase. For each phase marked done, verify the deliverable file exists and is non-empty. Report any gaps.

  4. Reconcile task tracking:

    • Are there items in tasks.md marked [x] but not in CHANGELOG.md?
    • Are there open GitHub issues with no matching task in tasks.md?
    • Are there tasks in tasks.md with no corresponding GitHub issue?
    • If any drift — propose fixes before making any changes.
  5. Run stale checks:

    • Any last_updated in wiki frontmatter older than 30 days? Flag.
    • Any source_file reference in wiki pointing to a file that no longer exists? Flag.
    • Any entity page mentioned in 3+ sources but with its own last_updated older than the newest source? Flag.
    • Any [[wikilink]] pointing to a non-existent page? Flag.
  6. Run CI gates locally (if configured):

    • Privacy grep: grep -r "<real_username>" site/ wiki/ must be empty
    • Link check: python3 -m llmwiki lint-docs (when implemented)
    • Performance budget: build time <15s, HTML <50MB
  7. Write the report. Structure:

    ## Health report: <project> — <date>
    
    **Phase:** <current phase> (<status>)
    
    ### ✅ Green
    - <things that are fine>
    
    ### ⚠️ Yellow
    - <minor drift that should be fixed soon>
    
    ### ❌ Red
    - <broken gates that block progress>
    
    ### Next action
    1. <specific thing the user should do next>
    2. <second thing if any>
    
  8. Offer to fix green issues automatically. For yellow/red, ask before touching anything.

Hard rules

  1. Never silently update _progress.md. Always explain what changed and why.
  2. Never close a GitHub issue without the user saying so.
  3. Never overwrite tasks.md without showing the diff first.
  4. Respect the framework's "one PR per concern" rule — if a fix spans multiple files, surface it as multiple suggestions, not one big patch.
  5. Never auto-bump versions. Version bumps are a launch activity, not maintenance.

Related skills

  • self-learn — when a new pattern emerges during maintenance, pipe it to self-learn to potentially add it to the framework.
  • llmwiki-query — when you need context from past sessions on why a decision was made.
  • llmwiki-lint — for wiki-specific lint work (overlaps with step 5).

Keep looking

Skills are one crate of 627,801. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.