メニュー
Systematic book curation pipeline. Evaluates every chapter on 8 dimensions, identifies overlaps, and produces a scored report with actionable recommendations.
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
SOC 職業分類に基づく
| name | curate-book |
| description | Systematic book curation pipeline. Evaluates every chapter on 8 dimensions, identifies overlaps, and produces a scored report with actionable recommendations. |
| allowed-tools | ["Read","Edit","Write","Bash","Grep","Glob","Task","AskUserQuestion"] |
Systematic book curation pipeline. Evaluates every chapter on 8 dimensions, identifies cross-chapter overlaps, and produces a scored report.
Modes:
check - Phase 1 only (fast automated metrics, no LLM)chapter <name> - Single chapter evaluationreport - Regenerate report from cached data (Phase 4 only)Run automated metrics collection and generate a metrics-only report (fast, no LLM):
.venv\Scripts\python.exe -u scripts/curation/generate-curation-report.py --metrics-only
Read and summarize _analysis/curation-report.md.
Regenerate the report from existing cached data (Phase 4 only):
.venv\Scripts\python.exe -u scripts/curation/generate-curation-report.py
Read and summarize _analysis/curation-report.md.
Evaluate a single chapter:
.venv\Scripts\python.exe -u scripts/curation/generate-curation-report.py --chapter <name> --metrics-only
Read the chapter content and the metrics from _analysis/curation-metrics.json. Read OUTLINE.md for book context. Then score the chapter on all 8 dimensions using the rubric below and write the evaluation JSON to _analysis/curation-evaluations/<slug>.json.
.venv\Scripts\python.exe -u scripts/curation/generate-curation-report.py
.venv\Scripts\python.exe -u scripts/curation/generate-curation-report.py --metrics-only
Read _analysis/curation-metrics.json to get the chapter list. Read OUTLINE.md for book context.
Divide chapters into 3 batches. Launch up to 3 Task agents in parallel (subagent_type: "general-purpose"), each evaluating a batch.
Each agent instruction should include:
_analysis/curation-evaluations/<slug>.jsonThe agent should read each chapter, score it, and write the evaluation file.
After all Phase 2 agents complete, launch a single Task agent (subagent_type: "general-purpose") that:
_analysis/curation-evaluations/_analysis/curation-metrics.json for automated metrics_analysis/curation-cross-chapter.jsonThe JSON structure should be:
{
"overlaps": [
{"chapter_a": "href", "chapter_b": "href", "similarity": 0.0-1.0, "recommendation": "text"}
],
"narrative_flow": {
"gaps": ["description"],
"redundancies": ["description"],
"strongest_sequences": ["description"]
},
"page_budget": {
"total_pages": 780,
"overweight_chapters": ["list"],
"underweight_chapters": ["list"]
}
}
.venv\Scripts\python.exe -u scripts/curation/generate-curation-report.py
Read and summarize _analysis/curation-report.md to the user.
Does this chapter directly advance the book's goal of getting nations to sign a 1% treaty redirecting military spending to clinical trials?
| Score | Criteria |
|---|---|
| 5 | Core to the thesis; book is incomplete without it |
| 4 | Strongly supports the thesis; important but not essential |
| 3 | Tangentially supports; useful context but could be summarized |
| 2 | Weak connection; mostly interesting but not mission-critical |
| 1 | No clear connection to the 1% treaty goal |
How much does this chapter overlap with other chapters in content or argument?
| Score | Criteria |
|---|---|
| 5 | Completely unique content and argument |
| 4 | Minor overlap (<10%) with other chapters |
| 3 | Moderate overlap (10-25%); some repeated arguments |
| 2 | Significant overlap (25-50%); major sections duplicate other chapters |
| 1 | Mostly redundant; could be merged into another chapter |
Is this chapter in the right location within the book's narrative flow?
| Score | Criteria |
|---|---|
| 5 | Perfect placement; builds on prior chapters and sets up next ones |
| 4 | Good placement; minor reordering might help |
| 3 | Acceptable but not optimal; reader might be confused by position |
| 2 | Misplaced; belongs in a different part of the book |
| 1 | Severely misplaced; disrupts narrative flow |
Clarity, voice, professionalism. The book has a distinctive voice (sardonic, data-driven, urgent) that should be preserved.
| Score | Criteria |
|---|---|
| 5 | Excellent writing; engaging, clear, professional with personality |
| 4 | Good writing; minor improvements possible |
| 3 | Adequate; some sections unclear or inconsistent in voice |
| 2 | Below standard; needs significant editing |
| 1 | Poor quality; confusing, unprofessional, or voice-less |
Are claims sourced? Are calculations sound? Does it cite credible sources?
| Score | Criteria |
|---|---|
| 5 | Every claim cited; calculations verified; strong evidence chain |
| 4 | Most claims cited; minor gaps in evidence |
| 3 | Some unsourced claims; evidence present but incomplete |
| 2 | Many unsourced claims; weak evidence base |
| 1 | Mostly unsourced; assertions without evidence |
Is the length justified by the content value? Could the same points be made in fewer pages?
| Score | Criteria |
|---|---|
| 5 | Every page earns its place; tight, dense content |
| 4 | Mostly efficient; minor trimming possible |
| 3 | Some padding; 20-30% could be cut without losing value |
| 2 | Significant bloat; 30-50% could be cut |
| 1 | Severely bloated; core content is a fraction of page count |
Can a non-expert (interested general reader) follow this chapter?
| Score | Criteria |
|---|---|
| 5 | Crystal clear for general audience; technical concepts explained |
| 4 | Mostly accessible; occasional jargon explained in context |
| 3 | Somewhat technical; some sections may lose general readers |
| 2 | Quite technical; requires background knowledge |
| 1 | Impenetrable for non-experts |
Would a skeptical economist, policy analyst, or investor find this convincing?
| Score | Criteria |
|---|---|
| 5 | Publication-quality rigor; methodology sound, caveats stated |
| 4 | Strong rigor; minor methodological gaps |
| 3 | Moderate rigor; some hand-waving or missing sensitivity analysis |
| 2 | Weak rigor; assertions that would not survive peer review |
| 1 | No analytical rigor; pure advocacy without evidence |
Each chapter evaluation JSON (_analysis/curation-evaluations/<slug>.json):
{
"href": "knowledge/path/to/chapter.qmd",
"evaluated": "2026-02-14T12:00:00",
"scores": {
"D1_mission_necessity": 4,
"D2_uniqueness": 5,
"D3_placement": 4,
"D4_writing_quality": 4,
"D5_evidence_quality": 3,
"D6_page_efficiency": 4,
"D7_accessibility": 5,
"D8_policy_rigor": 3
},
"rationale": {
"D1_mission_necessity": "Core chapter explaining the treaty mechanism...",
"D2_uniqueness": "Only chapter covering this specific topic...",
"D3_placement": "Good position after problem statement...",
"D4_writing_quality": "Strong voice, clear structure...",
"D5_evidence_quality": "Most claims cited but missing source for...",
"D6_page_efficiency": "Tight at 7 pages, every section needed...",
"D7_accessibility": "Clear for general audience...",
"D8_policy_rigor": "Sound methodology but missing sensitivity..."
},
"recommendations": [
"Add citation for claim about X on line Y",
"Consider adding sensitivity analysis for Z"
]
}
CHS = weighted average of D1-D8 scores, scaled to 0-100.
| CHS | Classification | Recommended Action |
|---|---|---|
| 85-100 | GREEN | Keep as-is |
| 70-84 | YELLOW | Polish (specific fixes) |
| 50-69 | ORANGE | Significant work (merge/restructure/rewrite) |
| 0-49 | RED | Major decision (cut/demote/rewrite) |
Critique book prose in the author's voice using patterns learned from his own git edit history (not a hand-written rubric). Use when asked to "voice-check", "punch up", "make this sound like me / like Vonnegut / like Cunk", or to review whether a passage is too pompous / cute / self-aware. Catches what the regex scanner (voice-check.py, ~1.4% of real edits) misses.
Review and improve QMD chapter writing quality by replacing boring/consultant text with Wishonia's deadpan alien voice, consolidating redundancy, and reducing word count. Use when asked to fix voice, make content funnier, remove jargon, or enforce the book's writing style.
Spawns parallel agents to review all paperback QMD files. Reads variable-resolved previews, flags judgment-required issues with REVIEW comments.
Automated book maintenance. Runs health checks, prioritizes work, and systematically fixes issues. Use when you want Codex to autonomously improve the book.
Check GitHub Actions workflow status, diagnose failures, fetch logs, and fix common build issues. Use when builds fail, deploys break, or you need to investigate CI/CD problems.
Critically analyze a calculated parameter for academic rigor. Checks methodology, identifies weaknesses, suggests improvements for economist audiences.