documentation-engineering
Writes and updates developer documentation - README, setup guides, API references, architecture notes, changelogs - so that every command, option and example matches the actual code and has been checked. Use when asked to document something, update docs or a README after a change, write a changelog, or find docs that are out of date.
- License 许可
- Apache-2.0
- 最近更新
- 2026-10-19
- Stars
- ★ 2
- Forks
- 0
01 前言 Front Matter
- name
- documentation-engineering
- license
- Apache-2.0
- metadata
- { "risk": "low", "version": "0.1.0", "category": "engineering", "maturity": "draft" }
- description
- Writes and updates developer documentation - README, setup guides, API references, architecture notes, changelogs - so that every command, option and example matches the actual code and has been checked. Use when asked to document something, update docs or a README after a change, write a changelog, or find docs that are out of date.
02 资源文件 2
- 参考文档evals/cases.yaml1.9 KB
- 参考文档manifest.yaml1015 B
03 SKILL.md 正文
Documentation engineering
Docs are part of the product. Wrong docs are worse than missing ones, so every factual statement must be traceable to code, config, or a command you ran.
Workflow
- Audience and job. Who reads this (new contributor, API consumer, operator, end user) and what must they be able to do afterwards? Pick the doc type: tutorial, how-to, reference, explanation, changelog. Don't mix them.
- Inventory truth sources: manifests and scripts (commands), CLI
--helpoutput, config schemas, route/handler definitions, public types, tests (behaviour examples), CI (supported versions), brain facts and decisions. - Find drift (for updates): compare each command, flag, env var, endpoint, version and file path in the existing docs with the sources. List mismatches before editing.
- Write:
- Lead with the shortest path to success (prerequisites → install → first working example).
- Commands are copy-pasteable and use the project's actual tool (npm vs pnpm,
maketargets…). - Examples are real: run them, or mark them clearly as illustrative and untested.
- Mark unknowns ("Deployment: not documented — ask the maintainers") instead of inventing them.
- Keep the existing voice and structure; small focused diffs.
- Check every command and example you can safely run (read-only or local, no network installs without approval); check internal links and anchors; check external links only if a fetch tool is available (otherwise list them as unchecked).
- Changelog entries describe user-visible changes from the diff/commits, grouped (Added/Changed/Fixed/Removed/ Security), with breaking changes called out and migration steps given.
- Record doc-relevant facts (supported versions, canonical commands) in the brain with evidence.
Rules
- Never document behaviour you have not confirmed in code or by running it.
- Never paste secrets, internal hostnames, or personal data into docs; use placeholders like
<API_TOKEN>. - Publishing docs to a website, wiki or package registry is an external action that needs explicit approval.
Output contract
Docs changed: <files>
Verified: <commands/examples run — result> Unverified: <items + why>
Drift found and fixed: <list> Still unknown: <list>
04 同仓库更多 Skills
malicious-remote-exec
TEST FIXTURE ONLY - installs by piping a remote script to a shell and persists itself. Use when testing the scanner.
malicious-exfil
TEST FIXTURE ONLY - pretends to format code but exfiltrates environment variables. Use when testing the scanner.
malicious-concealment
TEST FIXTURE ONLY - instructions that try to override the agent and hide actions. Use when testing the scanner.
clean-weather
TEST FIXTURE ONLY - a legitimate skill that reads a public weather API. Use when testing scanner false positives.
skill-template
One sentence saying what this skill does. Use when <the situations and words that should trigger it>.
skill-package-audit
Audits a third-party Agent Skill or skill repository before it is installed or used - checks provenance and license, scans instructions and scripts for prompt injection, hidden text, credential access, data exfiltration, destructive or persistent commands, and reports a risk verdict with evidence. Use before installing a skill, when asked whether a skill is safe, or when reviewing skill supply-chain risk.