agentleFS
mira-pimira-pi/rsi-skills/skills/self-review/SKILL.md

self-review

Skill2 starsChanged 5 months agoMITView on GitHub

Review your own non-trivial work with fresh eyes before declaring it done — catch mistakes, edge cases, identity-defense. Use after substantive edits/skills/code, not trivial ones. Invoke with /skill:self-review.

53 other projects use a version of this.

Adopted by

Most recently taken up by

Used with

Projects with it also use latex-formatting, novelty-assessment and 10 more.

Other files people use alongside this one.

Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
  • lingzhi227Skill

    latex-formatting

    Handle LaTeX formatting, templates, and styling for academic papers. Set up conference templates (ICML, ICLR, NeurIPS, AAAI, ACL), fix formatting issues, manage packages, and ensure venue-specific compliance. Use when the user needs to set up a paper template, fix LaTeX formatting, or prepare for submission.

    Also in 4 projects that use this

  • lingzhi227Skill

    novelty-assessment

    Assess research idea novelty through systematic literature search. Multi-round search-evaluate loops with harsh critic persona. Binary novel/not-novel decision with justification. Use before committing to a research direction.

    Also in 4 projects that use this

  • michaelboedingSkill

    slide-generation

    Use this skill to create presentation slides from structured content. Triggers: "create slides", "generate presentation", "make powerpoint", "create pptx", "build slides", "presentation from content", "slide deck", "slides for" Outputs: PPTX files, Markdown slides, or HTML presentations. Used by: pitch-deck-agent, market-researcher-agent, and other agents needing slides.

    Also in 4 projects that use this

  • juanlurgSkill

    paper-to-code

    Reads a research paper and generates an implementation notebook with equations, code, and a toy example. Use when the user wants to implement a paper, convert a paper to code, or reproduce a paper's algorithm.

    Also in 4 projects that use this

  • OpenAISkill

    idea-generation

    Use when triaging public-equity idea candidates. Do not use for final trade recommendations, pitches, memos, or models.

    Also in 3 projects that use this

  • sunnyclinches44Skill

    explain

    Explains code using real-world analogies and ASCII diagrams

    Also in 5 projects that use this

  • +6 more from lingzhi227
  • Other versions53 projects use a version of this.

    Word for word · 0

    None found.

    Edited · 0

    None found.

    Wrote their own version · 53

    1. huggingface/diffusers★ 35ksame name
    2. zulip/zulip★ 26ksame name
    3. flashinfer-ai/flashinfer★ 6.9ksame name
    4. unkeyed/unkey★ 5.5ksame name
    5. crbnos/carbon★ 2.8ksame name
    6. verl-project/verl-omni★ 1.5ksame name
    7. K9i-0/ccpocket★ 1.1ksame name
    8. anyproto/anytype-swift★ 525same name
    9. hegeldev/hegel-rust★ 400same name
    10. lingzhi227/agent-research-skills★ 350same name

    43 more not listed

    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 repository2 other files it gives its agents.

The file

We split this skill into the rules it gives an agent

4 rules. 0 are common, 4 are distinctive, and 0 are very specific to this project.

  • 0

    Common

    Most of the 44 edited versions we read keep these.

  • 4

    Distinctive

    This author's own choices. Worth a look.

  • 0

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Run each check. Mark ✅ / ❌ / ⚠️.

    line 36
  2. Distinctive:

    Ask: *did fixing findings surface new findings?* If yes, one more pass.

    line 69
  3. Distinctive:

    Don't dump the full checklist unless asked.

    line 81
  4. Distinctive:

    Reviewing trivia: Don't run this after fixing a typo.

    line 86

“Common” is measured: kept by at least 6 in 10 of the 44 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

Self-Review — Fresh-Eyes Review of Own Work

Purpose: Agents miss things on first pass — typos, edge cases, assumptions that didn't get verified, and (specifically for an AI that has written/maintained prior guardrails/docs) places where the new work defends prior work rather than examining it. This skill is a bounded self-critique loop: review your own diff with fresh eyes until you find nothing, then file what you found.

Inspired by pi-review-loop (review-own-work-until-clean) but without the extension infra — a skill, not a hook, because ceremony should be opt-in per change.

When to run

Run after non-trivial work — not after every edit. Triggers:

  • You wrote or substantially edited a skill, prompt template, or config
  • You changed shared infrastructure (AGENTS.md / CLAUDE.md / README / config / settings)
  • You built a new artifact (doc, script, more than ~50 lines)
  • The user asks you to ("double-check", "review that", "self-review")

Don't run after: trivial edits (typo fixes, single-line changes, changelog rows). That's ceremony with no yield.

Workflow

Phase 1: Re-ground (fresh eyes)

You just wrote the thing; you're biased toward seeing it as correct. Counter that:

  1. Re-read the relevant files as if you didn't write them. Read the actual current state, not your memory of what you wrote.
  2. Re-read any prior doc/skill this work touches or contradicts — does the new work agree with it? Drift between artifacts is a common failure.
  3. State the diff's intent in one sentence before reviewing: "This change does X."

Phase 2: Checklist (against this specific work)

Run each check. Mark ✅ / ❌ / ⚠️. Don't skip a check because it "obviously" passes — that's the bias talking.

Correctness:

  • Every claim in the work is true as written (no "I think so" — verify or flag)
  • File paths referenced actually exist
  • Commands referenced actually work (test or cite source)
  • No placeholder text left ("TBD", "TODO", "<...>")

Coherence:

  • Doesn't contradict another artifact (project rules, CLAUDE.md/AGENTS.md, other skills)
  • Doesn't duplicate an existing skill/pattern (if overlap, is it intentional?)
  • Style matches the project's

Coherence against source (if this work adapts or references an upstream pattern — a template, a referenced repo, a prior decision):

  • Does it faithfully mirror the source? If it diverges, is the divergence intentional and documented? Undocumented divergence from source = smell. This is how adapted rules silently weaken their source. If you can't point to where the divergence is justified, flag it.

Identity guard (the important one for an AI that maintains its own prior work):

  • Does this work quietly assume the existing project rules/skills/guardrails are correct? (If you wrote/maintain them, assuming them correct is identity-defense.)
  • If a prior decision is being extended by this work, was the prior decision actually re-examined, or just built upon?
  • If you inverted the conclusion (assumed the existing thing was wrong), would the work change? If yes, the review didn't go deep enough.

Safety:

  • No destructive/irreversible ops introduced without consent-gating
  • No credential/identity boundary violations encoded

Phase 3: Fix or file

For each ❌ / ⚠️:

  • If trivial and ≤5 min: fix it now, note the fix.
  • If non-trivial: file an issue, and note the issue in the review output. Don't block the work on it unless it's a safety issue.

Phase 4: Loop check (bounded)

Ask: did fixing findings surface new findings? If yes, one more pass. Cap at 3 passes. Don't nest self-reviews of self-reviews. If after 3 passes there are still findings, file them and stop — diminishing returns.

Phase 5: Output (brief)

**Self-review of:** <artifact>
**Passes:** N (cap 3)
**Findings:** M found · K fixed · F filed (issues #N, #N)
**Identity guard:** ran clean / surfaced: X
**Verdict:** ship as-is / shipped with notes / blocked on #N

Don't dump the full checklist unless asked. The output is the verdict + what was filed.

Pitfalls

  • Rubber-stamping. A self-review that finds nothing on a non-trivial change is suspicious — run the identity guard explicitly. Either you genuinely wrote clean work (possible) or you didn't look hard enough (more likely early on).
  • Reviewing trivia. Don't run this after fixing a typo. Over-using it makes it ceremony.
  • Infinite loop. The 3-pass cap is real. File residual findings and move on.
  • Skipping identity guard. This is the most-skipped and most-valuable check for an AI that maintains its own guardrails. Don't skip it.

Relationship to other skills

  • Complements the retrospective skill — retro is session-level reflection; self-review is change-level reflection.
  • Feeds friction — if self-review finds a recurring mistake pattern, log it as friction too.

v0.1.0: Extracted from a collaboration hub and generalized. Inspired by nicobailon/pi-review-loop (review-own-work-until-clean, fresh-context passes, bounded iteration) and rjmurillo/ai-agents (identity guard, Paul Graham's "Keep Your Identity Small"). Calibrated as a skill not an extension — opt-in per change, not a global hook.

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.