agentleFS
QuarktexQuarktex/qcobrain/.claude/skills/review/SKILL.md

review

Skill4 starsChanged 4 months agoMITView on GitHub

Fast evolution loop — run every ~N turns or at the end of a working session. Scans the session for new facts, decisions, and learnings, then routes each candidate through cite-or-refuse (sourced → wiki/, unsourced-but-worth-keeping → inbox/, junk → dropped per the blocklist) and appends an entry to runtime/review-log.md. Proposes writes; never bypasses governance. Trigger on "review session", "review the last N turns", end of a working session, or ~every 10 turns. Documented cron-able later.

192 other projects use a version of this.

Adopted by

Most recently taken up by

Used with

Projects with it also use close-spec, daily-arxiv 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.
  • dzhngSkill

    close-spec

    When a spec under specs/<feature>/ is done shipping (a write-spec build, or any planned task that produced a spec), archive it to specs/done/ and rewrite it from a build-plan into a durable rationale — the why, the principles, the invariants — pointing back to the real code for the how. Use when implementation has landed and the plan no longer matches what shipped, or the user says a feature/spec is finished. Pairs with [write-spec](../write-spec/SKILL.md) (the plan this closes) and [review](../review/SKILL.md) (run before closing).

    Also in 5 projects that use this

  • LambenthanSkill

    daily-arxiv

    每日从 arXiv 拉取新论文,过滤相关性,auto-ingest 高优先级论文,检测 SOTA 更新

    Also in 5 projects that use this

  • dzhngSkill

    design-with-images

    Explore UI and visual design with image generation, then iterate the real implementation against the user-selected concept. Use when brainstorming visual directions, offering mockup variants, or implementing an approved generated design; do not stop at the mockups or the first code approximation.

    Also in 5 projects that use this

  • LambenthanSkill

    exp-eval

    实验判决门:Review LLM 独立评判实验结果 → 4 种判决路径 → 自动更新 claims confidence、ideas status、graph edges

    Also in 5 projects that use this

  • +5 more from Lambenthan
  • +3 more from dzhng

The file

We split this skill into the rules it gives an agent

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

  • 0

    Common

    Most of the 156 edited versions we read keep these.

  • 0

    Distinctive

    This author's own choices. Worth a look.

  • 3

    Very specific

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

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

    Blocklist first: Check references/blocklist.md.

    line 30
  2. Very specific:

    Every note this loop causes to be created carries origin: agent-generated (the write-origin flag) — so /curator may later manage it, while …

    line 39
  3. Very specific:

    Only this loop's own origin: agent-generated notes are loop-managed.

    line 65

“Common” is measured: kept by at least 6 in 10 of the 156 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

Inherits references/kernel.md (the provenance contract). Reads references/wiki-spec.md for schema.

What this is

The fast half of qcobrain's two-speed metabolism (the slow half is /curator). It does not invent knowledge — it harvests what the session already produced and routes each item through the same autonomy gate every write obeys. Fully governed: autonomy never bypasses provenance.

When to run

  • ~every N turns (default 10) during a long working session, and
  • at the end of any state-changing session.

In v1 you invoke it. It is cron-able later — see runtime/README.md.

Execution

1 — Harvest candidates

Re-scan the session since the last runtime/review-log.md entry for three things only: new facts about you / your business / the market, decisions made, and process learnings. List each as a one-line candidate. Ignore everything else (chatter, tool output, scaffolding).

2 — Route each candidate through cite-or-refuse

For EACH candidate apply the gate (the spine — see references/kernel.md; do not restate it):

  1. Blocklist first. Check references/blocklist.md. If it is a transient/env-specific error, a negative tool-capability claim ("X doesn't work"), one-off narrative chatter, a secret/PII, or speculation stated as fact → drop it. Never written, not even to inbox/.
  2. Sourced (a verbatim quote re-resolves in a sources/ file, or one can be landed now) → propose a wiki/ note and hand the actual write to /capture, which owns source-landing + hashing
    • the verbatim check. This loop proposes; /capture proves.
  3. Unsourced but worth keeping → inbox/<slug>.md with status: needs-source and a one-line note on what source would confirm it. Never smuggled into wiki/.

Every note this loop causes to be created carries origin: agent-generated (the write-origin flag) — so /curator may later manage it, while your user-directed and pinned: true notes stay immune.

3 — Decisions and learnings

  • A decision → append to decisions/log.md (date | decision | why) and, if it is a durable claim, capture a type: decision note via /capture.
  • A process learning → /learn (appends runtime/learnings.md). Don't also duplicate it into wiki/.

4 — Log the pass

Append to runtime/review-log.md (create it with a header if it doesn't exist yet), newest on top:

## <date> — review (turns <range>)
captured: <n> → wiki   quarantined: <n> → inbox   dropped: <n> (blocklist)
notes: [id, id]   inbox: [slug]   decisions: [n]   learnings: [n]
needs your approval: <big/identity-level claims held back>

5 — Report

One screen: what was captured, what was quarantined, what was dropped and why, and anything awaiting your judgment. Big or identity-level claims are proposed, not auto-canonized.

Guardrails (hard)

  • Extracts only what the session actually produced; never fabricates to fill the log.
  • The blocklist is checked before anything else — poison never reaches inbox/, let alone wiki/.
  • Big/identity-level claims need your approval (kernel §3); flag, don't auto-write them.
  • Only this loop's own origin: agent-generated notes are loop-managed; user-directed and pinned: true are untouchable.
  • Provenance still bottoms out in sources/; this loop never short-circuits /capture's verbatim check.

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.