docs-update
Skill1 starsChanged 6 months agoMITView on GitHub
Use when finishing a session in a repo with a multi-file doc set (any 3+ of README.md, CLAUDE.md, CONTEXT.md, ARCHITECTURE.md, DECISIONS.md, WORKLOG.md, TODO.md) and the docs are behind the code. Routes each delta to the right file: appends to WORKLOG, edits TODO, records decisions, adds CONTEXT terms, updates ARCHITECTURE. Triggers on the user saying "catch up the docs", "sync the docs", "update the docs", "doc catch-up", "bring the docs current", "the docs are stale", "what about docs", "wrap up", "log this work", or invoking `/docs-update`. Pairs with the `context-md` skill (the CONTEXT.md-specific worker docs-update delegates to for glossary edits). Skip in single-doc repos (README only) and in repos with no doc set yet (use `context-md` first).
- Commits and pushes
17 other projects use a version of this.
Adopted by
Most recently taken up by
- microsoft/veronatheir own version, first seen Nov 8, 2026
- buildomator/buildomatortheir own version, first seen Nov 4, 2026
- aka-kika/akakika-skillstheir own version, first seen Nov 1, 2026
- davidanthonyn/claude-skillstheir own version, first seen Oct 20, 2026
- harukikuri/spotnotetheir own version, first seen Oct 20, 2026
- jellydn/my-ai-toolstheir own version, first seen Oct 18, 2026
Other versions17 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 17
- microsoft/verona★ 3.7ksame name
- warpdotdev/oz-skills★ 825same name
- bagofwords1/bagofwords★ 463same name
- hiromaily/go-crypto-wallet★ 128same name
- jellydn/my-ai-tools★ 127same name
- TangleML/tangle-ui★ 104same name
- buildomator/buildomator★ 102same name
- harukikuri/spotnote★ 25same name
- aka-kika/akakika-skills★ 11same name
- Escoto/RustifyMyClaw★ 8same name
7 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 repository3 other files it gives its agents.
- context-mdSkill
- klaxonSkill
- open-source-skillSkill
The file
We split this skill into the rules it gives an agent
5 rules. 0 are common, 5 are distinctive, and 0 are very specific to this project.
0
Common
Most of the 16 edited versions we read keep these.
5
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
- Distinctive: line 21
After a one-line fix that doesn't shift docs (typo, lint, dependency bump).
- Distinctive: line 35
Don't guess shape from filename.
- Distinctive: line 47
Use the in-context session arc (what happened in chat) as primary source of intent.
- Distinctive: line 65
Before writing, show the user a one-line summary per doc:
- Distinctive: line 91
Do NOT push without explicit approval.
“Common” is measured: kept by at least 6 in 10 of the 16 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.
Files
docs-update
Catch up the per-repo doc set with what actually happened this session. Read commits, code delta, worklog, and the chat arc; update each doc by appending the right kind of change to the right file.
The whole point: every doc has a job. Don't pile updates into whichever doc is closest. Route the right change to the right file.
When to use
- End of a work session in a repo with 3+ of the canonical docs above.
- The user says "catch up the docs", "sync docs", "doc catch-up", or invokes
/docs-update.
When NOT to use
- Repo has only README.md (single-doc repo). Just update the README in place; no skill needed.
- Mid-session refactor without a meaningful arc yet. Wait for the work to land.
- After a one-line fix that doesn't shift docs (typo, lint, dependency bump).
Process
1. Inventory
ls *.md | grep -E '^(README|CLAUDE|CONTEXT|ARCHITECTURE|DECISIONS|WORKLOG|TODO)\.md$'Note which of the 7 are present. Other md files (CHANGELOG, CONTRIBUTING, etc.) are out of scope for this skill.
2. Read each one
Don't guess shape from filename. Read the actual file. Note its current structure and which section is the append target (e.g. WORKLOG has a today-dated section; TODO has Open and Recently shipped).
3. Survey session changes
git log --oneline <since-marker>..HEAD # commits this session
git diff --stat <since-marker>..HEAD # files touched
git status --short # uncommittedFor <since-marker>: prior session's last commit, or today's first commit. Fall back to "last 24h" if unsure (--since='24 hours ago').
Use the in-context session arc (what happened in chat) as primary source of intent. The git log gives the artifacts.
4. Per-doc decision rules
For each doc that exists, decide: append, edit, or skip.
| File | Append/edit when |
| WORKLOG.md | Always (if any code shipped). One paragraph under today's date heading. New ## YYYY-MM-DD section if no entry for today. |
| TODO.md | When new open items surfaced OR existing items were shipped. Strike with ~~~~ OR move to "Recently shipped" section. Add new items to "Open". |
| DECISIONS.md | Only when a new design decision passes the 3-gate: hard to reverse, surprising without context, real trade-off. Otherwise skip. Most commits don't warrant a new decision entry. |
| CONTEXT.md | When new domain vocabulary was introduced. New noun in the system, new verb in the workflow, new flagged ambiguity. NOT for general programming concepts. (Delegate to the context-md skill.) |
| ARCHITECTURE.md | When module map changed (new file in src/), phase count changed, new external integration added, or storage shape changed. |
| README.md | When operator-facing surface changed: new CLI flag, new env var, new install step, new run command. Architecture section update only if ARCHITECTURE.md changed and you want them mirrored. |
| CLAUDE.md | When a new convention was established that future-session-Claude needs to know. Rare. Most session learnings belong in a memory system, not CLAUDE.md. |
5. Preview
Before writing, show the user a one-line summary per doc:
docs-update preview:
WORKLOG.md: +1 entry under 2026-05-17 (slice B, link-mode, duration fix)
TODO.md: +1 open (non-Shorts handling), 0 struck
DECISIONS.md: no change
CONTEXT.md: +2 terms (Link-mode, Persona facts)
ARCHITECTURE.md: no change
README.md: no change
CLAUDE.md: no change
proceed? (y / edit which one / skip)On y, write each one. On edit X, ask the user for the correction. On skip, abort.
6. Commit
One commit covering all doc changes:
git add WORKLOG.md TODO.md CONTEXT.md # etc., only changed files
git commit -m "docs: catch up after <one-line session topic>"Do NOT push without explicit approval. This skill commits; pushing (especially to a deploy branch that auto-builds) is a separate decision the user makes.
Common mistakes
- Padding WORKLOG with trivia. One paragraph per session, summarizing the arc. The git log has the detail. Don't restate every commit message in prose.
- Decision-record inflation. "We chose X over Y" is a decision IF Y was a real alternative and the cost of reversing is real. Most refactors don't warrant a record.
- General programming terms in CONTEXT.md. "Timeout" is not domain vocabulary. "Bundle" (when your system has a specific Bundle concept) is.
- Duplicating content across files. WORKLOG is the diary, DECISIONS is the why, ARCHITECTURE is the shape, CONTEXT is the vocabulary, README is operator-facing, CLAUDE is agent-facing. They reinforce, they don't repeat.
- Pushing without approval.
git pushto a deploy branch is a separate ask. This skill commits; deploy-decisions are the user's.
Red flags (stop and reconsider)
- About to write more than one paragraph to WORKLOG. Probably padding. Trim.
- About to add a 5-bullet decision record. Probably belongs as a single line in WORKLOG or a single bullet in TODO instead.
- About to mention "Claude" or "agent" in WORKLOG. The diary is for humans; agent learnings go to a memory system.
- About to add CLAUDE.md entries that overlap with an existing memory entry. Memory is the right home for cross-repo learnings.
Quick reference
docs-update flow:
inventory → read each → survey delta → per-doc decide
↓
preview
↓
user: y
↓
write + commit
(no push)Related:
- The
context-mdskill: CONTEXT.md format and the glossary-maintenance worker docs-update delegates to. - Continuous-diary cadence: append WORKLOG entries as work happens, not batched at session end.
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.

