Inside Claude Code's deep-research Skill: 349 Lines, 3 Adversarial Verifier Votes per Claim
first_class_orgins · reddit · 2026-09-15
A developer reverse-engineers Claude Code's bundled deep-research skill (349 lines of JS, ported from "bughunter architecture") that produced a 27-source, 123-claim report in one run. Key takeaways: three isomorphic prompts (role/context/task/decision-criterion/format), hard rules like "default to refuted=true", tuning constants at the top, per-agent JSON schemas, prompts as functions, explicit pipeline/parallel orchestration with an intentional barrier, and defensive early exits — a blueprint for building robust multi-agent skills.
More from coding & agent
- User builds 20-bot Grok suite to track wins, prep 1:1s and close promises — NicoVerderosa · 2026-10-30
- LangChain CEO opens Interrupt London with keynote on why orgs should own their intelligence — LangChain · 2026-10-30
- Simlock: open-source control plane gives parallel coding agents their own iOS/Android simulators — Scobleizer · 2026-10-30
- YC experiment: Claude agents formed democracies while GPT agents formed hives — soumitrashukla9 · 2026-10-30
- Keyfleet launches open canvas where personal AI agents team up with a shared USDC treasury — seanwbren · 2026-10-30
- Page-based project management for Grok Bots: keep chats short, history on Pages — AiJohnAllen · 2026-10-30