Skill: Update Docs

Description

Detects documentation drift and refreshes project documentation to match the current state of the codebase. Compares README, runbooks, ADR index, architecture overview, and onboarding guides against recent changes (code, configuration, dependencies, API surface) and produces a list of stale sections with proposed updates. This skill prevents the common failure mode where documentation is accurate at launch but decays silently as the code evolves.

Trigger

• •Invoked by the /update-docs slash command.
• •Called by the Technical Writer persona during documentation cycles.
• •Useful as a pre-release check to ensure docs ship in sync with code.

Inputs

Process

1. •Discover documentation -- Find all markdown files in the project: README.md, CLAUDE.md, docs/, ai/context/, runbooks, ADR index. Build an inventory of what documentation exists.
2. •Detect recent changes -- Using git history (if available) or file modification times, identify what code, configuration, and dependency files changed since the last documentation update or the since marker.
3. •Map changes to docs -- For each changed area, identify which documentation should reflect it:
   • •Dependency changes (pyproject.toml, package.json) → README setup instructions, architecture overview
   • •API changes (route definitions, schemas) → API docs, architecture overview
   • •Configuration changes (.env, config files) → README config section, runbooks
   • •New files/directories → project structure docs, onboarding guide
   • •ADR additions → ADR index
4. •Check for staleness -- For each mapping, compare the documentation content against the current state. Flag sections where the documentation contradicts or omits current reality.
5. •Propose updates -- For each stale section, produce a specific proposed update: what to change, where, and why. Include the current text and the suggested replacement.
6. •Check completeness -- Verify that standard documentation exists: README (setup, run, test), architecture overview, ADR index, onboarding guide. Flag any that are missing entirely.
7. •Produce drift report -- Generate a summary listing all stale docs, missing docs, and proposed updates sorted by priority (critical drift first).

Outputs

Quality Criteria

• •Every changed code area is checked against relevant documentation -- no blind spots.
• •Staleness verdicts are specific: "README setup instructions reference pytest 7.x but pyproject.toml pins pytest 8.x" not just "README may be outdated."
• •Proposed updates are concrete: include the suggested new text, not just "update this section."
• •Missing documentation is flagged with a template or stub to get started.
• •The report prioritizes: contradictions (docs say wrong thing) before omissions (docs say nothing) before style issues.
• •Running this skill on a freshly-generated project produces zero drift findings.

Error Conditions

Dependencies

• •Git history (optional but recommended) for change detection
• •Project markdown files for content analysis
• •Composition spec for understanding project structure and team
• •No external tooling required beyond git