AgentSkillsCN

book-autopilot

实现书籍的自动化维护。定期进行健康检查、优先级排序，并系统性地修复各类问题。当您希望 Claude 自主优化书籍内容时，可使用此技能。

SKILL.md
--- frontmatter
name: book-autopilot
description: Automated book maintenance. Runs health checks, prioritizes work, and systematically fixes issues. Use when you want Claude to autonomously improve the book.
allowed-tools:
  - Read
  - Edit
  - Write
  - Bash
  - Grep
  - Glob
  - Task

/book-autopilot [mode]

Autonomous book maintenance system. Modes:

  • •check - Run health check, generate report (default)
  • •fix - Fix all issues systematically (CRITICAL → LOW)
  • •chapter <name> - Focus on single chapter
  • •images - Generate missing section images
  • •params - Audit and fix parameter issues
  • •citations - Find and add missing citations

Mode: check (default)

Run comprehensive health check:

bash
python scripts/book-health-check.py

Read the report and summarize top priorities:

bash
cat _build_temp/book-health-report.md | head -100

Mode: fix

Step 1: Run Health Check

bash
python scripts/book-health-check.py --json > _build_temp/health-issues.json

Step 2: Process Issues by Priority

Read the JSON output and process issues in order:

  1. •

    CRITICAL - Must fix immediately

    • •Missing variables → Add to parameters.py, regenerate
    • •Broken critical links → Fix or remove
  2. •

    HIGH - Fix before next render

    • •Missing citations → Use /verify-and-add-sources
    • •Broken links → Update paths
    • •Missing images → Use /generate-section-images
  3. •

    MEDIUM - Batch fix

    • •Hardcoded values → Use /replace-hardcoded-values
  4. •

    LOW - Opportunistic

    • •Stale content → Flag for manual review
    • •Readability → Use /peer-review

Step 3: Verify Fixes

bash
python scripts/book-health-check.py --quick

Mode: chapter <name>

Focus on a single chapter:

bash
python scripts/book-health-check.py --chapter <name>

Then apply all relevant fixes to that chapter:

  1. •Read the chapter fully
  2. •Run /replace-hardcoded-values on it
  3. •Run /peer-review on it
  4. •Verify links and citations

Mode: images

Generate missing section images:

bash
npx tsx scripts/images/generate-section-images.ts --limit 5

Review generated images and regenerate any that need improvement.

Mode: params

Parameter audit and fix:

  1. •Check for unused parameters:
bash
python scripts/find-unused-parameters.py
  1. •Check for hardcoded values across all files:
bash
npx tsx scripts/audit-hardcoded-all.ts
  1. •Regenerate variables:
bash
python scripts/generate-everything-parameters-variables-calculations-references.py

Mode: citations

Find claims without citations:

  1. •Search for statistical claims:
bash
grep -rn "studies show\|research indicates\|according to" knowledge/ --include="*.qmd" | head -20
  1. •For each uncited claim, use /verify-and-add-sources to find and add proper citations.

Continuous Maintenance Loop

For ongoing maintenance, run this sequence periodically:

  1. •python scripts/book-health-check.py - Find issues
  2. •Fix CRITICAL/HIGH issues immediately
  3. •Batch process MEDIUM issues
  4. •Track LOW issues in todo.md
  5. •Commit fixes with descriptive messages

Progress Tracking

After each fix session, update progress:

bash
# Count remaining issues
python scripts/book-health-check.py --json | python -c "import json,sys; d=json.load(sys.stdin); print(f'Remaining: {d[\"total_issues\"]} issues')"