AutoResearchClaw Research Extract

**Research Date:** 2026-03-18 **Target Repo:** https://github.com/aiming-lab/AutoResearchClaw **Focus:** Novel patterns worth stealing for CLAW mesh integration AutoResearchClaw is a 23-stage autonomous research pipeline that transforms a single idea into conference-ready papers. Three patterns stand out as immediately valuable: 1. **4-layer citation verification** — prevents LLM hallucination in references (arXiv → DOI → OpenAlex → Semantic Scholar fallback) 2. **Cross-run lesson extraction** — JSONL append-only failure log → LLM skill generation → prompt overlay injection (25% retry reduction) 3. **PIVOT/REFINE/PROCEED autonomy** — autonomous decision loops with versioned rollbacks and pivot limits The rest is solid engineering but not novel: multi-source literature search (OpenAlex/S2/arXiv), sandbox execution with AST validation, hardware detection, quality gates with template ratio heuristics. | Source | Endpoint | Rate Limit | Query Method | |--------|----------|------------|--------------| | **OpenAlex** | `https://api.openalex.org/works` | 10K/day | `search` param + `filter` for dates, `select` for fields, `mailto` for polite pool | | **Semantic Scholar** | `https://api.semanticscholar.org/graph/v1/paper/search` | 1 req/s (free), 10 req/s (paid) | `query` param + `year` filter + `fields` selection | | **arXiv** | `https://export.arxiv.org/api/query` | 1/3s | `search_query=all:{query}` + Atom XML parsing |