agentleFS
X-T-E-RX-T-E-R/kiki/.agents/skills/kiki-docs-catchup/SKILL.md

kiki-docs-catchup

Skill53 starsChanged 14 days agoMITView on GitHub

Catch Kiki maintainer or user documentation up after a coherent implementation candidate, or execute an explicit documentation-debt task. Use when behavior is already implemented and tested and the task is to classify documentation impact, establish source and test evidence, choose canonical paths, update bilingual views, or prepare exact mechanical documentation wiring. Do not use during ordinary early implementation or for read-only repository questions.

The file

We split this skill into the rules it gives an agent

4 rules. About 0 are common, 2 are distinctive, and 2 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈2

    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
  1. Very specific:

    Read references/trigger-matrix.md when the boundary is ambiguous or several documentation tools appear applicable.

    line 22
  2. Distinctive:

    Choose one fact home and its views: Prefer the owning package contract or generated projection.

    line 30
  3. Very specific:

    Verify and report: Run the smallest claim-matched checks plus the repository's pnpm docs:check-governance command.

    line 33
  4. Distinctive:

    Stop with the missing evidence or decision and the smallest next handoff.

    line 46

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 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.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Kiki docs catch-up

Bring documentation up to the boundary that the implementation candidate is approaching. This workflow starts after coherent behavior and tests exist, or when the user explicitly assigns documentation debt; it does not delay exploration or early coding.

Trigger matrix

SituationTriggerResult
A tested candidate is preparing for stable reuse, owner handoff, default-on, or public releaseYesClassify impact and update the due views
An explicit task names existing documentation debtYesReconstruct evidence, then update the bounded views
worker_lite needs exact, already-decided copy, insertion, navigation, or link-shape workYesValidate the semantic packet, then freeze a mechanical packet
Feature implementation is still exploratory or tests do not establish the claimed behaviorNoContinue implementation and testing
A user asks what current Kiki docs sayNoUse the read-only kiki-docs skill

Read references/trigger-matrix.md when the boundary is ambiguous or several documentation tools appear applicable.

Workflow

  1. Fix the candidate boundary. Name the branch, commit range, or diff and the approaching boundary. Do not broaden the implementation or invent release timing.
  2. Gather independent evidence. For each material claim, read the owning implementation and a claim-matched test whose assertion would fail if the behavior changed. Read existing owner and audience views only after the behavior evidence is clear.
  3. Classify impact and completion separately. Choose exclusive none, or select every affected category from generated, maintainer, and user. Then record completed or deferred; none pairs with completed. Generated work is never deferrable. Every deferred maintainer or user entry retains all applicable categories and its exact view, one accountable owner, and a target no later than the earliest applicable boundary: stable reuse or owner handoff for maintainer work, or default-on or public release for user work.
  4. Assign lineage. Mark each claim inherited, adapted, Kiki-only, or unknown. Documentation or accepted product history establishes lineage; current code and tests establish behavior but do not infer provenance.
  5. Choose one fact home and its views. Prefer the owning package contract or generated projection. Update links and audience views instead of repeating package inventories. Use references/semantic-packet.md to record claims, audiences, lineage, evidence, canonical paths, and acceptance before editing.
  6. Apply the due work. Keep generated contracts exact in the current candidate. Complete maintainer views before stable reuse or owner handoff. Complete both docs/en/ and docs/zh/ views before default-on or public release. Follow the documentation directory's governing instructions and the runtime Public-Facing Content rules for reader-facing prose.
  7. Delegate only frozen mechanics to worker_lite. The semantic owner writes and approves all meaning-bearing prose, including final target-language copy, directly or through the existing translate-docs workflow before filling references/mechanical-packet.md. worker_lite may only place frozen copy, insert exact snippets, and apply exact navigation or link-shape mappings. It receives no authority to translate, invent claims, select pages, classify lineage, or change source maps.
  8. Verify and report. Run the smallest claim-matched checks plus the repository's pnpm docs:check-governance command. Inspect both locale views and every generated projection changed. Report proof limits: the structural checker does not establish semantic correctness.

For post-release changelog curation, hand off to sync-changelog. For translation of an already-approved page and structure, use translate-docs within the frozen semantic boundary.

Stop conditions

  • The implementation candidate or its claim-matched test is missing.
  • The due boundary cannot be determined and different answers would change which audience must be complete.
  • Source and tests disagree, or existing views assert conflicting behavior.
  • A mechanical request requires choosing claims, pages, lineage, or mappings that the semantic packet did not freeze.
  • A mechanical packet omits the final approved target-language copy that worker_lite would place.
  • Work would cross the supplied write scope, expose non-public values, or publish externally.

Stop with the missing evidence or decision and the smallest next handoff. Do not fill a semantic gap with polished prose.

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.