update-docs

Scans project documentation for staleness and inaccuracies, then updates it to reflect the current codebase. Uses a read-only scanner subagent to assess docs and a writer subagent to apply approved changes. Optionally scoped to a specific area via arguments.

Safety Notice

This listing is imported from skills.sh public index metadata. Review upstream SKILL.md and repository scripts before running.

Copy this and send it to your AI assistant to learn

Install skill "update-docs" with this command: npx skills add zackbart/motif/zackbart-motif-update-docs

Documentation Updater

Update this project's documentation to accurately reflect the current state of the code. Delegates discovery to a scanner subagent and updates to a writer subagent, with user approval in between.

Argument Parsing

Parse $ARGUMENTS for an optional focus area:

  • If $ARGUMENTS is empty, scan all documentation in the project.
  • If $ARGUMENTS contains a focus area (e.g. "README", "api docs", "CLAUDE.md", "installation steps", "architecture"), scope the scan to that area. Pass the focus to the scanner subagent so it narrows its search.

Store the focus (or lack thereof) for use in the briefings below.


Step 1 — Scan

Spawn the docs-scanner subagent. Pass it a briefing containing:

  1. Project root: run ls at the project root and include the output
  2. Focus area: what the user wants scoped (or "all documentation" if no args)
  3. Key context: language/framework if obvious from the root listing

The scanner will return a structured report of what documentation exists, what's outdated, and what's missing. Display the full report to the user.


Step 2 — User approval

Present the scanner's findings and ask the user what to act on:

Documentation scan complete. [Summary of findings]

Which updates should I apply?

  • all — apply everything the scanner found
  • pick — let me choose specific items
  • none — stop here, I just wanted the audit

Wait for the user's response. If they pick specific items, confirm the list before proceeding.


Step 3 — Update

For each approved update, spawn the docs-writer subagent. Pass it a briefing containing:

  1. The file to update: full path
  2. What needs to change: the specific finding from the scanner
  3. Current file contents: read the file and include it so the writer has the full context (not just the stale section)
  4. Codebase truth: any relevant source files the writer needs to reference for accuracy — include paths so it can read them

If there are multiple independent files to update, you may spawn multiple writer subagents in parallel.

Each writer returns the changes it made. Collect all results.


Step 4 — Summary

After all writers complete, present a file-by-file summary:

  • File path
  • What was changed (brief)
  • What was left alone

If any writer encountered something it couldn't resolve (ambiguous source of truth, conflicting information), flag it for the user.

Source Transparency

This detail page is rendered from real SKILL.md content. Trust labels are metadata-based hints, not a safety guarantee.

Related Skills

Related by shared tags or category signals.

General

optimize-prompt

No summary provided by upstream source.

Repository SourceNeeds Review
General

second-opinion

No summary provided by upstream source.

Repository SourceNeeds Review
Coding

dev

No summary provided by upstream source.

Repository SourceNeeds Review