MicrosoftKnown publishermicrosoft/DocAtlas-Harness/docatlas/skills/review/SKILL.md

review

Skill21 starsChanged 9 days agoMITView on GitHub

Use this skill to recall previously saved analysis notes by a natural-language query. An auxiliary LLM looks over all saved notes' cards (note_id, page_refs, found text) and selects only the relevant ones, which Review then returns in full. Triggers include "review the notes on topic X", "do we have evidence about Y in our notes", "pull the partisan-split notes", "summarize what we've found about ethics". Always prefer Review over re-reading pages when the information might already be in a note — it saves an expensive page-read round-trip. Requires at least one Note has been written in this session.

We split this skill into the rules it gives an agent

7 rules. 0 are common, 5 are distinctive, and 2 are very specific to this project.

0
Common
Most of the 161 edited versions we read keep these.
5
Distinctive
This author's own choices. Worth a look.
2
Very specific
Tied to their project's names, files or tools. Probably skip.

Common
Distinctive
Very specific

Distinctive:
Before producing the final answer — ask Review for everything you learned that bears on the user's question.
line 21
Distinctive:
When you think you *might* have answered a sub-question earlier but don't want to scroll back through turns.
line 23
Distinctive:
Before a new Search — pulling prior notes on the topic first prevents re-asking queries you've already answered.
line 25
Very specific:
No notes have been saved yet (Review will just tell you so).
line 30
Very specific:
If no notes match, selected_note_ids is empty and text says so.
line 52
Distinctive:
After a few turns with saved notes:
line 56
Distinctive:
Note → Review: write then recall. The natural pairing.
line 74

“Common” is measured: kept by at least 6 in 10 of the 161 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

Review — recall saved notes by query

The second half of the Note / Review pair. Note writes findings down; Review pulls them back out. The mechanism: build a short "note card" per saved analysis (id + page refs + found text), let an auxiliary LLM select which cards match the query, and return the full rendered bodies of those notes.

When to use

Before producing the final answer — ask Review for everything you learned that bears on the user's question.
When you think you might have answered a sub-question earlier but don't want to scroll back through turns.
Before a new Search — pulling prior notes on the topic first prevents re-asking queries you've already answered.

When NOT to use

No notes have been saved yet (Review will just tell you so).
The information you need is clearly not in any note — Review only looks at saved notes, never at pages.

Arguments

Field	Required	Notes
`query`	yes	Focused recall question. Phrase it like a search query.

Output

{
  "text": "Selected 2 note(s) for query 'partisan ethics': …rendered notes…",
  "query": "partisan ethics",
  "selected_note_ids": [1, 3],
  "candidate_note_ids": [1, 2, 3],
  "rationale": "notes 1 and 3 mention partisan splits on ethics-related items…"
}

If no notes match, selected_note_ids is empty and text says so.

Example

After a few turns with saved notes:

Direct calls require HARNESS_SESSION_FILE; see ../README.md.

uv run --locked python docatlas/skills/review/scripts/run.py \
    --query "what did we learn about partisan splits on presidential ethics?"

Failure modes

aux LLM call failed — the Azure endpoint/credentials for the aux LLM aren't available. Check HARNESS_AUX_LLM_* / AZURE_OPENAI_* env vars.
Empty selection on a query you expect to hit — re-phrase the query more specifically or widen it.

Tips for combining with other skills

Note → Review: write then recall. The natural pairing.
Review → final answer: just before composing the answer, do one Review with the user's question as the query.

Download now (ZIP)

Other versions191 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 191
garrytan/gstack★ 138ksame name
ansible/ansible★ 71ksame name
docling-project/docling★ 70ksame name
traefik/traefik★ 65ksame name
ClickHouse/ClickHouse★ 51ksame name
temporalio/temporal★ 24ksame name
tisfeng/Easydict★ 15ksame name
juicedata/juicefs★ 15ksame name
elie222/inbox-zero★ 13ksame name
gitkraken/vscode-gitlens★ 9.9ksame name
181 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.
Used alongsideProjects with it also use daily-arxiv, exp-eval and 10 more.
Other files people use alongside this one.
- 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
- 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
- SethGammonSkill
  ascii-diagram
  Generate perfectly aligned ASCII diagrams — architecture, flow, sequence, box-and-arrow. Uses a programmatic character-grid approach so alignment is guaranteed by math, not token prediction. Includes post-render verification.
  Also in 4 projects that use this
- DAIR-AutoResearchSkill
  exp-pilot-eval
  Pilot result evaluation — read pilot results, apply success criteria, update idea page (pilot_result, failure_reason if failed), generate PILOT_VERDICT_REPORT. Called by /ideate Phase 5 after /exp-pilot-run.
  Also in 4 projects that use this
- DAIR-AutoResearchSkill
  exp-pilot-run
  Pilot experiment execution — read Pilot Spec YAML, write pilot code, run experiment(Confirm with the user before operation and require the applicant to conduct manual inspection), return results. Called by /ideate Phase 5. Does NOT modify wiki pages or judge pass/fail.
  Also in 4 projects that use this
- dzhngSkill
  audit-choices
  Audit the choices an implementing agent made, not its diff — a pure decision audit that traces the session's history into a choices ledger, changes no code, and never blocks an unsupervised run. Working code still embeds architecture the user never chose; surface it because future work inherits it. Use when the user wants to review the decisions the AI made on their behalf, before merging or committing AI-implemented work, when integrating a delegated subagent's pass, or when a fix "works" but might be a point fix.
  Also in 4 projects that use this
- +5 more from Lambenthan
- +1 more from DAIR-AutoResearch
In this repository4 other files it gives its agents.
- web/AGENTS.mdAGENTS.md
- noteSkill
- readSkill
- searchSkill

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

Reviews can't be read right now.

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.

Check your setup, free

7 rules. 0 are common, 5 are distinctive, and 2 are very specific to this project.

Review — recall saved notes by query

When to use

When NOT to use

Arguments

Output

Example

Failure modes

Tips for combining with other skills

Word for word · 0

Edited · 0

Wrote their own version · 191

daily-arxiv

exp-eval

ascii-diagram

exp-pilot-eval

exp-pilot-run

audit-choices

Does it work?

Your trail

Now do this for your own.