agentleFS
tbhbtbhb/vale-ai-tells/.claude/skills/codex-fix-prose/SKILL.md

codex-fix-prose

Skill118 starsChanged 37 days agoMITView on GitHub

Clear vale and cspell findings on a settled document through an isolated Codex writer and an independent reviewer. Use when a draft's prose gates are red or repeated lint findings would pollute the parent context. Takes a target and its complete lint command.

The file

We split this skill into the rules it gives an agent

6 rules. About 0 are common, 5 are distinctive, and 1 is very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈5

    Distinctive · estimate

    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. Distinctive:

    Keep findings and retries out of the parent context.

    line 9
  2. Very specific:

    Stop on BLOCKED. On PROCEED, create a plan with update_plan: baseline, writer, mechanical verification, independent review, final lint.

    line 15
  3. Distinctive:

    Only the delegated writer may edit the target, and the target is the only file it may edit.

    line 33
  4. Distinctive:

    Never change lint configuration, dictionaries, hooks, or suppression comments.

    line 34
  5. Distinctive:

    Don't infer clean from silent output unless preflight proved the target has prose rules.

    line 35
  6. Distinctive:

    Don't expose full lint findings in the parent response unless the workflow blocks and the operator needs a decision.

    line 36

≈ “Common” is an estimate: at least 1 in 5 of the 372,429 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

Orchestrate a prose fix

Keep findings and retries out of the parent context. Parse the request into a target and complete lint command, then run:

bash .agents/skills/codex-fix-prose/scripts/preflight.sh <target> <lint command>

Stop on BLOCKED. On PROCEED, create a plan with update_plan: baseline, writer, mechanical verification, independent review, final lint.

Record the baseline before delegation:

bash .agents/skills/codex-fix-prose/scripts/check-suppressions.sh --baseline <target>

Save the absolute state: path that command prints. It identifies this workflow and records its target. Until cleanup, the script refuses a second workflow on the same target. Pass the state to every delegated writer and reviewer in this run.

Delegate codex-write-prose-fix with fork_turns: "none". Give it only the repository root, target, workflow state, lint command, and instruction to follow that skill. Wait for its terminal report. Never edit the target in the parent.

Run check-suppressions.sh --verify <state> <target>. Any finding means the run isn't clean. Delegate codex-review-prose-fix with fork_turns: "none", and give it the repository root, target, and workflow state. If it returns changes required, save those findings in <state>/notes, delegate the writer again with the same minimal prompt, and repeat verification and review. Bound this writer/reviewer loop at three review rounds.

Finally run the complete lint command once in the parent. Always run check-suppressions.sh --cleanup <state> before returning once no delegated agent needs the state, even when the outcome is unresolved or blocked. Report PROSE: CLEAN only when the writer reported clean, the suppression check is silent, the independent reviewer returned REVIEW: CLEAN, and the final lint command passes. Otherwise return the short unresolved or blocked report. Mark each plan item as it completes.

Safety rules:

  • Only the delegated writer may edit the target, and the target is the only file it may edit.
  • Never change lint configuration, dictionaries, hooks, or suppression comments.
  • Don't infer clean from silent output unless preflight proved the target has prose rules.
  • Don't expose full lint findings in the parent response unless the workflow blocks and the operator needs a decision.

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.