agentleFS
HenryQWHenryQW/skills/issue-blueprint/SKILL.md

issue-blueprint

Skill1 starsChanged 4 months agoApache-2.0View on GitHub

Turn a rough multi-issue plan into a user-approved spec and dependency-aware GitHub issue graph. Use when asked to plan a multi-issue feature or create child issues plus a tracker and one final_check. Do not use for a standalone spec or glossary edit.

The file

We split this skill into the rules it gives an agent

8 rules. About 0 are common, 6 are distinctive, and 2 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈6

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 2

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Read only relevant project instructions, specs, glossary, repository state, issue state, and any supplied Repo Surveyor report.

    line 23
  2. Distinctive:

    Do not create or switch branches, commit, or run target-repository validation while the bundle is provisional.

    line 33
  3. Very specific:

    After PASS, show the exact spec path, target repository, issue titles, dependency waves, labels, and non-goals.

    line 35
  4. Distinctive:

    If approval is withheld or the graph changes, keep the bundle provisional and return to review.

    line 36
  5. Distinctive:

    Only ambiguity, incorrectness, unsafe behavior, untestability, poor slicing, avoidable interference, or improper sequencing are blockers.

    line 46
  6. Distinctive:

    Check requested labels, then run:

    line 50
  7. Distinctive:

    Prefer the fewest cohesive, independently verifiable, non-interfering vertical children.

    line 67
  8. Very specific:

    Use the schema and validation ownership in references/issue-plan.md as the sole issue-shape contract.

    line 68

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Issue Blueprint

Turn a rough multi-issue plan into an approved GitHub issue graph without implementing it.

Memory

Direct invocation owns $agent-memory; nested invocation defers it to the caller. Capture only accepted durable product, domain, or architecture decisions.

Workflow

  1. Route before planning:
    • one known actionable issue -> $issue-workbench #<issue>;
    • clear multi-issue feature -> continue;
    • standalone spec or glossary edit -> stop;
    • missing repository, spec location, or product behavior -> discover authoritative facts, then ask only for unresolved product decisions.
  2. Read only relevant project instructions, specs, glossary, repository state, issue state, and any supplied Repo Surveyor report. A Surveyor handoff is evidence, not a plan: Issue Blueprint alone owns issue-plan JSON, slicing, dependencies, child validation, adversarial review, and publication. Use project terminology and test domain relationships with concrete scenarios.
  3. Draft the provisional bundle:
    • implementation handoff spec in the project's spec location;
    • glossary changes only for changed terms and an ADR only for a surprising, hard-to-reverse trade-off;
    • issue-plan JSON conforming to references/issue-plan.md, including non-goals and any dropped_findings;
    • locally rendered issue graph. Do not create or switch branches, commit, or run target-repository validation while the bundle is provisional. The caller owns the single planning commit after finalization.
  4. Run blocker rounds with exactly one read-only adversarial reviewer per round. Pass exact artifact and repository paths. The first round must trace every acceptance criterion to concrete evidence and enforce the issue-plan slicing rules. Resolve discoverable facts from authoritative sources; report inaccessible sources as access blockers and ask the user only for material product decisions. Record answers, update the bundle, and re-render before reviewing only changed or unresolved material.
  5. After PASS, show the exact spec path, target repository, issue titles, dependency waves, labels, and non-goals. Publish only after explicit approval of that exact graph; earlier plan approval is not publication approval.
  6. If approval is withheld or the graph changes, keep the bundle provisional and return to review. Otherwise publish and finalize as below.

Reviewer contract

The reviewer remains read-only, does not question the user, and returns PASS or at most five material blockers ordered by dependency impact and severity:

severity | kind | evidence | issue | recommended resolution

Only ambiguity, incorrectness, unsafe behavior, untestability, poor slicing, avoidable interference, or improper sequencing are blockers. Recorded user decisions remain authoritative absent cited contradictory evidence. Missing concrete evidence for an acceptance criterion is an untestability blocker.

Publish and finalize

Check requested labels, then run:

python3 <skill_dir>/scripts/publish_issue_plan.py plan.json --repo OWNER/REPO --label enhancement --label ready-for-agent --verify

--verify self-tests, renders, publishes children then tracker, writes .context/issues/numbers.json, and prints only the absolute working directory and Use $shipyard #<parent> handoff. Do not duplicate successful verification with gh issue view.

The publisher checkpoints each issue and binds its state to the approved plan and repository. On partial failure, fix the cause and rerun the same command with --resume; never republish from a changed plan or repository.

Approval withheld or publication incomplete leaves the spec provisional. After complete publication, set an existing spec lifecycle marker to exactly Approved and published; do not invent one. Retry failed local finalization without republishing. Only then may the caller create the single planning commit.

Graph rules

  • Prefer the fewest cohesive, independently verifiable, non-interfering vertical children; keep non-goals visible and avoid unrequested compatibility or future-proofing.
  • Use the schema and validation ownership in references/issue-plan.md as the sole issue-shape contract.
  • The tracker contains the complete graph and explicit waves. Publish blockers before blocked work so bodies can use real issue numbers.
  • final_check is the only multi-child integration check. Return defects to the implementation child that owns the failed criterion.

Output discipline

Pass paths instead of pasted artifacts, keep progress phase-level, and never summarize unchanged documents. The main agent owns evidence, blocker disposition, user decisions, and edits.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.