{
  "markdown": "# recursive-research\n\n> Claude Code **plugin & skill** for recursive research up to PhD level on any topic — science, tech, business, arts, humanities. Source tiering, loop auto-regulation, disk checkpointing, and WDM + Munger inversion for autonomous decisions.\n\n**Version:** 2.2.0 · **License:** [MIT](LICENSE) · **Author:** Joseph Huayhualla ([@Anjos2](https://github.com/Anjos2))\n\n---\n\n## What it does\n\nYou give it a **research seed** (a topic) and the skill:\n\n1. Asks for mode (`web` / `local` / `mixed`), local paths if applicable, priority/excluded sources, and a cycle cap.\n2. Identifies **3-5 seed threads** applying [WDM (Weighted Decision Matrix) + Munger Inversion](#wdm--munger-inversion).\n3. Detects available MCPs (Firecrawl, Context7, WebFetch, WebSearch) and prioritizes by speed and quality.\n4. **Iterates in auto-regulated cycles** — each cycle picks the least-covered thread, selects sources, investigates, consolidates.\n5. Tiers every source into **Tier 1 / 2 / 3 / Rejected** with transparent criteria.\n6. Saves **disk checkpoints** every cycle — survives context compaction.\n7. Closes when the **5-criteria PhD fitness function** is met, or upon hitting the cycle cap.\n8. Asks if you want to keep going. **Research can be infinite.**\n\n---\n\n## Why it's different\n\n| Feature | How it solves it |\n|---|---|\n| Works across **any domain** | Generic source tiering (papers, academic books, official archives, raw data), not code-only |\n| **Rejects garbage sources** automatically | Explicit criteria: no author, data-less marketing, SEO spam, unsupervised AI content |\n| **Survives context limits** | Per-cycle disk checkpoint + `--resume` mode for new sessions |\n| **Self-critical** | Munger inversion applied to the consolidated knowledge: what do I not know? what bias do my sources share? what's missing? |\n| **Asks before assuming** | Full Phase 0 interrogation of the user |\n| **Transparent** | Every non-trivial autonomous decision runs WDM + Munger and shows the reasoning |\n\n---\n\n## Installation\n\nThis repo is a **Claude Code marketplace** containing one plugin (`recursive-research`). The easiest way to install is through Claude Code's built-in plugin manager.\n\n### Option A — Via marketplace (recommended)\n\nInside Claude Code, run:\n\n```\n/plugin marketplace add Anjos2/recursive-research\n/plugin install recursive-research\n```\n\nThe plugin appears in `/plugin` → Installed. Invoke it with `/recursive-research:recursive-research`.\n\n### Option B — Via the official Plugin Directory (pending approval)\n\nAnthropic maintains a [central Plugin Directory](https://claude.com/plugins). Once approved there (submission in progress), install becomes a one-liner without the marketplace add step:\n\n```\n/plugin install recursive-research\n```\n\n### Option C — As a standalone skill (minimal, no plugin manager)\n\nIf you prefer to bypass the plugin ecosystem, copy only the skill markdown:\n\n**Linux / macOS:**\n\n```bash\ngit clone https://github.com/Anjos2/recursive-research.git\nmkdir -p ~/.claude/skills/recursive-research\ncp recursive-research/plugins/recursive-research/skills/recursive-research/SKILL.md ~/.claude/skills/recursive-research/\n```\n\n**Windows (PowerShell):**\n\n```powershell\ngit clone https://github.com/Anjos2/recursive-research.git\nNew-Item -ItemType Directory -Force -Path \"$HOME/.claude/skills/recursive-research\"\nCopy-Item recursive-research/plugins/recursive-research/skills/recursive-research/SKILL.md \"$HOME/.claude/skills/recursive-research/\"\n```\n\nInvoked without namespace as `/recursive-research`. Verify with `/help` inside Claude Code.\n\n---\n\n## Usage\n\n### Standard invocation\n\n```\n/recursive-research:recursive-research     # if installed as plugin\n/recursive-research                        # if installed as standalone skill\n```\n\nThe skill guides you interactively. Answer in natural language.\n\n### Resume a paused research\n\n```\n/recursive-research:recursive-research --resume <slug>\n```\n\n`<slug>` = kebab-case name of the topic (e.g., `episodic-memory-humans`).\n\n### List saved research\n\n```\n/recursive-research:recursive-research --list\n```\n\n---\n\n## Usage examples by domain\n\n| Domain | Suggested seed |\n|---------|------------------|\n| Neuroscience | \"Mechanisms of episodic memory in humans\" |\n| Philosophy | \"Modern application of Stoic philosophy\" |\n| Music | \"Minimalism in 20th-century music\" |\n| Business | \"B2B SaaS monetization models in 2025\" |\n| History | \"Fall of the Western Roman Empire: economic causes\" |\n| Biology | \"CAR-T cell immunotherapy against cancer\" |\n| Technology | \"Hexagonal architecture in microservices\" |\n| Law | \"European AI Act and its extraterritorial impact\" |\n\n---\n\n## \"PhD level\" criterion\n\nThe skill only declares PhD when **all 5 criteria** are met:\n\n1. **Coverage ≥80%** across all seed threads\n2. **≥3 Tier-1 sources per thread**\n3. **New-finding saturation ≤5%** for 3 consecutive cycles\n4. **Munger inversion applied** to the knowledge (what I don't know, what sources contradict, what biases exist)\n5. **≥3 explicit cross-thread connections** between different threads\n\nIf any criterion fails, it **does not declare PhD** and keeps iterating — or asks for confirmation upon hitting the cycle cap (default 20, configurable).\n\n---\n\n## Source tiering\n\n| Tier | Qualifies | WDM weight |\n|------|-----------|------------|\n| **1** | Peer-reviewed papers · academic books · official standards (W3C, RFC, ISO, WHO) · primary archives · official datasets | 5 |\n| **2** | Official repos · blogs from citable authors · recorded conferences · Wikipedia with references · reports with methodology | 3 |\n| **3** | Blogs with citations to T1/T2 · high-voted forum answers with sources · recorded interviews with identifiable experts | 2 |\n| **Reject** | No author · data-less marketing · SEO spam · tutorials without sources · unsupervised AI content | 0 |\n\nEvery consulted source is logged in a per-tier file for later audit.\n\n---\n\n## Pre-loaded seed sources\n\nThe skill auto-suggests reliable sources by domain. Examples:\n\n- **Science**: arXiv, Semantic Scholar, Google Scholar, Connected Papers, OpenReview\n- **Medicine**: PubMed, Cochrane Library, WHO, ClinicalTrials.gov\n- **Humanities**: JSTOR, SSRN, Project MUSE\n- **Code**: GitHub, Context7, RFCs, W3C specs\n- **Data**: World Bank, OECD Data, Our World in Data, Pew Research\n- **Art/culture**: Europeana, Google Arts & Culture, Internet Archive, Project Gutenberg\n\nThe user can add or reject any before starting.\n\n---\n\n## WDM + Munger inversion\n\nDecision frameworks applied at each non-trivial autonomous step:\n\n- **WDM (Weighted Decision Matrix)** — enumerate 3+ viable alternatives, criteria with weights, 1-5 scoring, compared totals.\n- **Munger inversion** (Charlie Munger via Jacobi) — ask inverted about the winning option: *\"how would it fail? what bias does it have? what am I ignoring?\"*\n\nThe skill applies both to:\n- Select seed threads\n- Select sources per cycle\n- Decide when to close the research\n- Validate the consolidated knowledge at the end\n\nReference: [Charlie Munger's essay on mental inversion](https://fs.blog/inversion/).\n\n---\n\n## Recommended MCPs\n\nThe skill detects what you have and uses them in this order:\n\n1. **[Firecrawl MCP](https://github.com/mendableai/firecrawl-mcp-server)** (highly recommended) — AI-optimized scraping\n2. **[Context7 MCP](https://github.com/upstash/context7)** — official library docs\n3. **WebSearch + WebFetch** (built-in Claude Code) — universal fallback\n4. **Chrome DevTools MCP** — only when content requires real JS execution; generally **too slow** for large-scale research\n\n---\n\n## Generated files\n\nIn `memoria/investigaciones/<slug>/` of the active project:\n\n- `estado.md` · `hilos.md` · `hallazgos.md`\n- `fuentes-tier-1.md` · `fuentes-tier-2.md` · `fuentes-tier-3.md` · `fuentes-rechazadas.md`\n- `ciclo-01.md`, `ciclo-02.md`, ..., `ciclo-N.md` (checkpoints)\n- `sintesis.md` · `acciones.md` · `gaps.md` (upon closing)\n\n**If `memoria/` doesn't exist in the project, the skill creates it** (notifying the user) — it's an explicit dependency.\n\n---\n\n## Repository structure\n\n```\nrecursive-research/                              ← the repo is a Claude Code marketplace\n├── .claude-plugin/\n│   └── marketplace.json                        ← marketplace manifest (declares the plugin)\n├── plugins/\n│   └── recursive-research/                     ← the plugin (named same as the marketplace)\n│       ├── .claude-plugin/\n│       │   └── plugin.json                     ← plugin manifest\n│       └── skills/\n│           └── recursive-research/\n│               └── SKILL.md                    ← the skill instructions (the actual content)\n├── LICENSE                                      ← MIT\n├── README.md                                    ← this file\n├── PRIVACY.md                                   ← privacy policy (no data collection)\n└── .gitignore\n```\n\n---\n\n## Contributing\n\nIssues and PRs welcome. If you improve a criterion, add a tier, find a new anti-pattern, or want support for more domains: open a PR.\n\n### Roadmap\n\n- [ ] Native integration with reference managers (Zotero, Mendeley)\n- [ ] Export research to academic formats (LaTeX, BibTeX)\n- [ ] Collaborative mode — multiple agents investigating threads in parallel\n- [ ] Automatic source-quality metrics (h-index, journal impact factor)\n- [ ] Support for PDFs behind legal paywalls\n\n---\n\n## GitHub topics\n\n`claude-code` · `claude-code-skill` · `claude-code-plugin` · `ai-agent` · `research-tool` · `recursive-research` · `knowledge-management` · `weighted-decision-matrix` · `mental-models`\n\n---\n\n## Privacy\n\n`recursive-research` does not collect any user data. Everything runs locally on your machine. See [PRIVACY.md](PRIVACY.md) for the full policy.\n\n## License\n\n[MIT](LICENSE) — use, modify, distribute freely. Just keep the copyright notice.\n\n---\n\n## Acknowledgments\n\n- **Charlie Munger** — for \"Invert, always invert\" (via Carl Jacobi)\n- **Claude Code team** — for the skill and plugin format\n- **Anthropic** — for the model that makes this possible\n\nIf this skill was useful, consider giving the repo a ⭐.\n",
  "bytes": 10061,
  "sha": "e3d2d29b2a72b8bec904ff350a21df2357ead9e7c911ce44801c194a8d467029",
  "repo_slug": "anjos2/recursive-research",
  "fonte": "repo",
  "truncated": false,
  "api": "https://agentalog.com/api/listings/plg_anjos2_recursive_research_recursive_rese_1635887b/readme"
}