TokenRollAI

update-doc

"Use when user says 'update docs', 'sync documentation', 'refresh docs', or after code changes need documenting. Updates llmdoc system based on recent code changes."

TokenRollAI 452 50 Updated 4mo ago
GitHub

Install

npx skillscat add tokenrollai/cc-plugin/update-doc

Install via the SkillsCat registry.

SKILL.md

/update-doc

This skill updates the project's llmdoc documentation to reflect recent code changes.

Pre-fetched Context

  • Llmdoc structure: !find llmdoc -name "*.md" 2>/dev/null | head -50
  • Recent changes (3 commits): !git diff HEAD~3..HEAD --stat 2>/dev/null | head -30
  • Affected files: !git diff HEAD~3..HEAD --name-only 2>/dev/null | head -30
  • Llmdoc index: !cat llmdoc/index.md 2>/dev/null | head -50

Actions

  1. Step 1: Analyze Changes

    • If $ARGUMENTS is provided, use it as the description of what changed.
    • Otherwise, analyze the pre-fetched git diff to understand what changed.
  2. Step 2: Identify Impacted Concepts

    • Map changed files to llmdoc concepts:
      • Config files (.eslintrc, etc.) → reference/coding-conventions.md
      • Auth files → relevant architecture docs
      • New features → may need new docs
    • Create a list of impacted documents.
  3. Step 3: Update Documents

    • For each impacted document, use recorder agent with this prompt:
      Task: Update documentation for <concept_name>.
      Changes: <relevant git diff summary>
      Mode: content-only
      Principle: Use fewest words necessary.
  4. Step 4: Synchronize Index

    • After all updates complete, invoke a single recorder agent to:
      • Re-scan /llmdoc directory
      • Ensure index.md is consistent and up-to-date
      • Mode: full
  5. Step 5: Report

    • Summarize all documents created/updated/deleted.

Update Principles

  • Minimality: Use fewest words necessary
  • Accuracy: Based on actual code, not assumptions
  • No Code Blocks: Reference with path/file.ext:line format
  • LLM-Friendly: Write for machine consumption