universal-research-orchestratorlisted

# Universal Research Orchestrator ## Overview Domain-agnostic 5-phase research pipeline: SEARCH (parallel lanes) → SYNTHESIZE → PRODUCE → VALIDATE → ARCHIVE. **Three rules:** 1. NEVER implement before gathering context. 2. NEVER stop at first result. 3. SYNTHESIZE before acting. ## Pipeline at a Glance ``` Phase 0: Calibrate scope — domain, depth, output target Phase 1: Search (parallel) — 8–12 lanes, deep navigation, autonomous lanes Phase 2: Synthesize — cross-reference, verify, gap-fill Phase 3: Produce — structured output with citations Phase 4: Validate — completeness, citations, examples, confidence Phase 5: Archive — curl all source URLs, build _index.json Todo tracking — real-time across all phases ``` --- ## Phase 0: Scope Calibration Confirm before searching. Default to **comprehensive** when the user says "most complete", "exhaustive", or "long horizon". - **Domain?** security / codebase / framework / market / general → ask if unclear - **Depth?** summary / comprehensive / exhaustive → default comprehensive - **Output target?** file path, format, citation style → ask if unclear - **Include basics?** Default no — focus on niche, non-obvious, edge cases --- ## Phase 1: SEARCH — Maximize Coverage ### 1.1 Lane Decomposition Break the target into **8–12 orthogonal research lanes**, each a standalone parallel agent. - **Orthogonal**: each lane targets a distinct surface; redundancy wastes agents. - *