agentleFS
Ahsan-ireAhsan-ire/rag-pipeline-v1/.claude/skills/plan-gate/SKILL.md

plan-gate

Skill0 starsChanged 35 days agoMITView on GitHub

Pre-implementation quality gate — makes the design an artifact, then runs two independent fresh-context critiques (plan-auditor subagent + Codex second-vendor) and reconciles the findings. Nothing gets implemented until this gate says READY.

4 other projects use a version of this.

Adopted by

Most recently taken up by

  • Other versions4 projects use a version of this.

    Word for word · 0

    None found.

    Edited · 0

    None found.

    Wrote their own version · 4

    1. NUX-Design/claude-skills-fable-opus★ 24same name
    2. eranw2000/multi-phase-skills-framework★ 10same name
    3. AgriciDaniel/plan-gate★ 9same name
    4. pengxiaozi/dsh-plugins★ 6same name

    How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.

  • In this repository4 other files it gives its agents.

The file

We split this skill into the rules it gives an agent

3 rules. 0 are common, 2 are distinctive, and 1 is very specific to this project.

  • 0

    Common

    Most of the 3 edited versions we read keep these.

  • 2

    Distinctive

    This author's own choices. Worth a look.

  • 1

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Second critique — Codex, second vendor: Use the codex MCP tool if present, otherwise Bash:

    line 26
  2. Distinctive:

    READY: no unresolved blockers; every finding has a disposition.

    line 42
  3. Distinctive:

    Do not begin implementing anything during this gate, including "quick fixes" the critics suggest.

    line 48

“Common” is measured: kept by at least 6 in 10 of the 3 edited versions of this file we could read. “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

You are running the plan gate on $ARGUMENTS. The point of this gate is that the critics never saw the reasoning that produced the plan — do not brief them on intent, history, or "what we meant". Pointers only.

  1. Make the design an artifact. The plan must exist as a file before review: either a phase section of IMPLEMENTATION_PLAN.md or a docs/designs/NNN-<slug>.md document (copy docs/designs/TEMPLATE.md; next NNN in sequence). If the plan so far lives only in conversation, write the artifact first and confirm the file path with the user before proceeding — the artifact, not the chat, is what implementation will be checked against.
  2. First critique — plan-auditor subagent. Spawn the plan-auditor subagent (Agent tool). Pass: the plan file path, the acceptance criteria location, and nothing else. Its contract is a numbered failure-mode list with severities.
  3. Second critique — Codex, second vendor. Use the codex MCP tool if present, otherwise Bash: codex exec --sandbox read-only "Adversarially review <plan file> against IMPLEMENTATION_PLAN.md and CLAUDE.md: real bugs, missing steps, spec divergence, weak acceptance criteria. Cite file:line. Where a fix is small and mechanical, include a proposed unified diff in the finding (text only — you cannot apply it). Do NOT read data/, chroma_db/, or held-out eval files." Pointers only — NEVER paste corpus text, chunk contents, or held-out questions into the prompt (CLAUDE.md hard rule). If the codex CLI is not installed, record "second-vendor leg SKIPPED — codex unavailable" in the gate report; a silently missing leg is a gate failure mode of its own.
  4. Reconcile. Merge the two finding lists, dedupe, then verify each finding against the actual plan/code before accepting it — critics can be wrong or out of scope. Classify each: ACCEPT (revise the plan), REBUT (state the evidence), or OUT-OF-SCOPE (say where it goes instead, e.g. the cut list). Append the table to the design artifact under a ## Review heading — the dispositions are part of the design record.
  5. Verdict.
    • READY: no unresolved blockers; every finding has a disposition. Implementation may begin, against the artifact.
    • REVISE: numbered revision list. Apply revisions to the artifact (with user approval if they change scope), then re-run from step 2 — revised plans get fresh critics, not a sympathetic re-read.

Do not begin implementing anything during this gate, including "quick fixes" the critics suggest. The gate produces a verdict and an updated artifact, nothing else.

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.