agentleFS
dikamilodikamilo/dx-workflow/skills/dx-review-triage/SKILL.md

dx-review-triage

Skill8 starsChanged 4 months agoMITView on GitHub

Triage findings from a review report and apply the fixes you choose.

The file

We split this skill into the rules it gives an agent

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

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈1

    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:

    If applying a fix would contradict something the plan or a standard states elsewhere, stop and say so — don't silently pick a side.

    line 46
  2. Very specific:

    Every finding in the resolved report has a Resolution: other than PENDING — or you've stopped partway, which is fine, since re-running …

    line 50

≈ “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

dx-review-triage

Turn a review report's findings into decisions — and, when you say so, into edits. Every review (dx-review) only analyzes and reports; this is the one place that acts on a finding, one finding at a time, only on your confirmation.

Guard. Resolve <container-id> under context/changes/ or context/efforts/. If it resolves under context/archive/ instead, refuse: archived work is done. If reviews/ is missing or holds no report, point at /dx-review <policy-id> <container-id>.

1 — Resolve which report

A report lives at reviews/<policy-id>.md, one per Policy, always the current review because a re-run overwrites it.

  • A policy ID given → reviews/<policy-id>.md. If the file is missing, name it and list the reports that exist.
  • No policy ID → exactly one report → use it; several → ask which, since each can carry open findings at once.

A legacy reviews/impl-review.md or reviews/plan-review.md is never read and doesn't count as a report: if one is all there is, say it's superseded and point at /dx-review implementation <container-id> or /dx-review plan <container-id> respectively.

Load first

The report's schema — the finding/Resolution format, the resume rule, the never-commit rule — is context/config/templates/review-report.md (missing → /dx-init).

2 — Walk findings in order

Resume at the first finding with Resolution: PENDING, document order. Load what that finding's Location names — a plan section → plan.md; a file:line → that file plus enough surrounding code or git log to judge the fix — and nothing for findings you skip. Show its title, location, detail, and suggested fix, then ask:

  • Fix now — apply the suggested fix (or a variant you propose); show the before/after edit first, apply on confirmation.
  • Fix differently — ask what they'd prefer instead, apply that.
  • Skip — leave it for now, move on.
  • Accept risk — leave it, record the one-line reason.
  • Record as lesson — hand the finding's title and detail to /dx-lesson's entry shape and tell the user to confirm there. Never append to foundation/lessons.md yourself — that file has exactly one writer.

Immediately rewrite that finding's Resolution: line in place per the schema. If an edit you already applied for an earlier finding also resolves a later one, mark the later finding FIXED too, pointing at the same edit — don't manufacture a redundant second edit or commit.

3 — Commit what the fix touched, never the report

Decide per applied fix, from the files it changed:

  • Touched files outside context/ (shipped code, docs, config) → commit that fix on its own, the Conventional Commit shape dx-implement uses: fix(<container-id>): <finding title> (review). Stage only those files.
  • Touched only context/ artifacts (plan.md, frame.md, …) → don't commit. Those are working state the user commits on their own schedule.
  • The report itself is never committed.

On disagreement

If applying a fix would contradict something the plan or a standard states elsewhere, stop and say so — don't silently pick a side.

Done when

Every finding in the resolved report has a Resolution: other than PENDING — or you've stopped partway, which is fine, since re-running resumes from the next pending one. Print a one-line tally and stop, no auto-chain:

Triaged <report file>: <n> fixed, <n> skipped, <n> accepted, <n> dismissed
Next: /dx-review <policy-id> <container-id>   — re-review after fixes
  or: <each line of the Policy's ## Next>

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.