review-workspace
Skill0 starsChanged 2 months agoMITView on GitHub
Periodic maintenance pass over the workspace: run lint, surface stale and orphaned pages, verify and promote learnings, refresh work/now.md prompts, and run the evaluation questions. Use weekly or when asked to "tidy", "review", "audit", or "check freshness" of the workspace.
In this repository19 other files it gives its agents.
- agents/AGENTS.mdAGENTS.md
- AGENTS.mdAGENTS.md
- areas/AGENTS.mdAGENTS.md
- company/AGENTS.mdAGENTS.md
- knowledge/AGENTS.mdAGENTS.md
- scripts/AGENTS.mdAGENTS.md
- work/AGENTS.mdAGENTS.md
- agents/CLAUDE.mdCLAUDE.md
- areas/CLAUDE.mdCLAUDE.md
- CLAUDE.mdCLAUDE.md
- company/CLAUDE.mdCLAUDE.md
- knowledge/CLAUDE.mdCLAUDE.md
- scripts/CLAUDE.mdCLAUDE.md
- work/CLAUDE.mdCLAUDE.md
- answer-company-questionSkill
- capture-learningSkill
- import-knowledgeSkill
- record-decisionSkill
- research-questionSkill
The file
We split this skill into the rules it gives an agent
3 rules. About 0 are common, 1 is distinctive, and 2 are 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.
2
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 13
Lint: Run python3 scripts/lint.py --strict.
- Very specific: line 30
Now: Read work/now.md. Ask the owner whether the focus, finish line, and next decisions are still current.
- Distinctive: line 36
Decisions: Check "Revisit when" conditions on approved decisions.
≈ “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.
Files
Review the workspace
This pass keeps the workspace trustworthy. It proposes changes; the owner approves anything that touches approved knowledge.
Steps
- Lint. Run
python3 scripts/lint.py --strict. Fix mechanical errors (frontmatter, broken relative links, numbering) directly. - Stale pages. Lint lists pages past
review_by. For each, read it and decide: still true (bumpreview_by, andupdatedonly if content changed), needs an update (draft the change, keepstatusunchanged, flag for owner), or no longer true (status: superseded, link replacement, considerarchive/). Never silently extend a review date on a page you did not read. - Learnings. For each
knowledge/learnings/page:draft: can it be verified from its evidence? If yes, setapprovedand tell the owner. If it has been draft through two reviews, resolve it: promote, correct, or supersede.approved: has it proved durable? Fold it into the page named in "Promote to", then set the learning tosupersededwith a link.
- Orphans. Find pages under
company/,areas/,work/,knowledge/that noREADME.mdor other page links to. Link them from the right index or archive them. - Now. Read
work/now.md. Ask the owner whether the focus, finish line, and next decisions are still current. Do not change it yourself beyond formatting. - Projects. For each
work/projects/page, check the tracker or ask the owner for status. Close finished projects followingwork/projects/README.md. - Decisions. Check "Revisit when" conditions on approved decisions. If a condition is met, tell the owner; do not reopen the decision yourself.
- Evaluations. Answer every question in
agents/evals.mdusing onlyanswer-company-question. Record pass/fail and what you fixed asknowledge/learnings/YYYY-MM-DD-eval-run.md. A failure is a workspace bug: fix the page or the registry, not the answer. - Security. Confirm no secrets, credentials, or restricted personal data
entered the repository since the last review (
git log -pon the period, plus lint's secret scan).
Report
One message: what was fixed directly, what is proposed and awaits the owner (with paths), eval results, and the learnings promoted. Keep it under thirty lines.
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
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.

