{"id":19535,"plugin_id":"plugins_6a99cda269b0819180d65893e1de7811","kind":"skill","collection_source":null,"comparison_source":null,"observed_at":"2026-09-30T23:15:38.227Z","digest":"c29d01625ca8cec50784a4d983bc9f7ae8663112d105914555a7ebcf07146874","against":null,"payload":{"name":"deep-research","description":"Use when a task needs enterprise-grade research across many sources with citation tracking and verification. Triggers on \"deep research\", \"comprehensive analysis\", \"research report\", \"compare X vs Y\", \"analyze trends\", or \"state of the art\". Not for simple lookups, debugging, or questions answerable with 1-2 searches.","included_files":[{"relative_path":".gitignore","size_in_bytes":248},{"relative_path":"reference/continuation.md","size_in_bytes":4229},{"relative_path":"reference/html-generation.md","size_in_bytes":3061},{"relative_path":"reference/methodology.md","size_in_bytes":13811},{"relative_path":"reference/quality-gates.md","size_in_bytes":5623},{"relative_path":"reference/report-assembly.md","size_in_bytes":3468},{"relative_path":"reference/weasyprint_guidelines.md","size_in_bytes":5814},{"relative_path":"requirements.txt","size_in_bytes":579},{"relative_path":"scripts/citation_manager.py","size_in_bytes":6018},{"relative_path":"scripts/md_to_html.py","size_in_bytes":9965},{"relative_path":"scripts/research_engine.py","size_in_bytes":16469},{"relative_path":"scripts/source_evaluator.py","size_in_bytes":9492},{"relative_path":"scripts/validate_report.py","size_in_bytes":12398},{"relative_path":"scripts/verify_citations.py","size_in_bytes":15468},{"relative_path":"scripts/verify_html.py","size_in_bytes":7800},{"relative_path":"templates/mckinsey_report_template.html","size_in_bytes":10394},{"relative_path":"templates/report_template.md","size_in_bytes":13178},{"relative_path":"tests/fixtures/invalid_report.md","size_in_bytes":288},{"relative_path":"tests/fixtures/valid_report.md","size_in_bytes":3331}],"skill_md_contents":"---\nname: deep-research\ndescription: Use when a task needs enterprise-grade research across many sources with citation tracking and verification. Triggers on \"deep research\", \"comprehensive analysis\", \"research report\", \"compare X vs Y\", \"analyze trends\", or \"state of the art\". Not for simple lookups, debugging, or questions answerable with 1-2 searches.\n---\n\n# Deep Research\n\n## Core Purpose\n\nDeliver citation-backed, verified research reports through a structured pipeline with source credibility scoring, evidence persistence, and progressive context management.\n\n**Autonomy Principle:** Operate independently. Infer assumptions from context. Only stop for critical errors or incomprehensible queries.\n\n---\n\n## Decision Tree\n\n```\nRequest Analysis\n+-- Simple lookup? --> STOP: Use the available current web-search tool\n+-- Debugging? --> STOP: Use standard tools\n+-- Complex analysis needed? --> CONTINUE\n\nMode Selection\n+-- Initial exploration --> quick (3 phases, 2-5 min)\n+-- Standard research --> standard (6 phases, 5-10 min) [DEFAULT]\n+-- Critical decision --> deep (8 phases, 10-20 min)\n+-- Comprehensive review --> ultradeep (8+ phases, 20-45 min)\n```\n\n**Default assumptions:** Technical query = technical audience. Comparison = balanced perspective. Trend = recent 1-2 years.\n\n---\n\n## Workflow Overview\n\n| Phase | Name | Quick | Standard | Deep | UltraDeep |\n|-------|------|-------|----------|------|-----------|\n| 1 | SCOPE | Y | Y | Y | Y |\n| 2 | PLAN | - | Y | Y | Y |\n| 3 | RETRIEVE | Y | Y | Y | Y |\n| 4 | TRIANGULATE | - | Y | Y | Y |\n| 4.5 | OUTLINE REFINEMENT | - | Y | Y | Y |\n| 5 | SYNTHESIZE | - | Y | Y | Y |\n| 6 | CRITIQUE | - | - | Y | Y |\n| 7 | REFINE | - | - | Y | Y |\n| 8 | PACKAGE | Y | Y | Y | Y |\n\n---\n\n## File locations (read this first)\n\nChoose behavior based on the current execution surface:\n\n| Context | Helper execution | Report destination |\n|---|---|---|\n| Codex with local file and shell access | Resolve the installed `deep-research` skill directory from this loaded `SKILL.md`; call helpers from its `scripts/` directory | A user-approved workspace path such as `./research-output/[topic]-YYYYMMDD/` |\n| ChatGPT or another surface without local shell access | Do not claim that bundled scripts ran | Return the report in conversation or through the supported downloadable-file workflow |\n\nNever write into a home directory or external system without the user's authorization.\n\n## Execution\n\n**On invocation, load relevant reference files** (relative paths from this SKILL.md, which work on both surfaces):\n\n1. **Phase 1-7:** Load [methodology.md](./reference/methodology.md) for detailed phase instructions\n2. **Phase 8 (Report):** Load [report-assembly.md](./reference/report-assembly.md) for progressive generation\n3. **HTML/PDF output:** Load [html-generation.md](./reference/html-generation.md)\n4. **Quality checks:** Load [quality-gates.md](./reference/quality-gates.md)\n5. **Long reports (>18K words):** Load [continuation.md](./reference/continuation.md)\n\n**Templates:**\n- Report structure: [report_template.md](./templates/report_template.md)\n- HTML styling: [mckinsey_report_template.html](./templates/mckinsey_report_template.html)\n\n**Scripts:** on a local execution surface, resolve the installed skill directory first and use absolute package-relative paths:\n\n- `python3 [skill-dir]/scripts/validate_report.py --report [path]`\n- `python3 [skill-dir]/scripts/verify_citations.py --report [path]`\n- `python3 [skill-dir]/scripts/md_to_html.py [markdown_path]`\n\nIf local execution is unavailable, apply the same checks manually and disclose that the bundled validators did not run.\n\n---\n\n## Output Contract\n\n**Required sections:**\n- Executive Summary (200-400 words)\n- Introduction (scope, methodology, assumptions)\n- Main Analysis (4-8 findings, 600-2,000 words each, cited)\n- Synthesis & Insights (patterns, implications)\n- Limitations & Caveats\n- Recommendations\n- Bibliography (COMPLETE - every citation, no placeholders)\n- Methodology Appendix\n\n**Output files** — use the destination contract above. Produce:\n- Markdown (primary source)\n- HTML when the current surface can create it\n- PDF when a supported PDF workflow or WeasyPrint is available\n\n**Quality standards:**\n- 10+ sources, 3+ per major claim\n- All claims cited immediately [N]\n- No placeholders, no fabricated citations\n- Prose-first (>=80%), bullets sparingly\n\n---\n\n## Common Mistakes\n\n- **Running deep/ultradeep for a question one search answers** — burns 10-45 minutes for no added confidence. Walk the decision tree first; simple lookup means STOP.\n- **Citing a major claim from a single source** — fails the 3+ sources standard and ships unverified assertions. Triangulate, or downgrade the claim and say so in Limitations.\n- **Leaving placeholder or uncited entries in the bibliography** — `validate_report.py` rejects it, and a client who checks one dead citation distrusts the whole report. Every [N] must resolve.\n- **Writing bullet-heavy sections** — violates the prose-first (>=80%) contract and reads as notes, not analysis. Convert findings to narrative; bullets only for genuine lists.\n- **Skipping `verify_citations.py` before packaging** — fabricated or dead links reach the reader. Run it on every report, every time.\n\n## When to Use / NOT Use\n\n**Use:** Comprehensive analysis, technology comparisons, state-of-the-art reviews, multi-perspective investigation, market analysis.\n\n**Do NOT use:** Simple lookups, debugging, 1-2 search answers, quick time-sensitive queries.\n"},"changes":[],"summary":"First saved snapshot. No earlier version is available for comparison.","summary_kind":"deterministic","summary_metadata":{}}