Open-source protocol for AI agents doing longform industry research: brief gates, state files, source/claim discipline, staged drafting, review loops, hard stops, gotchas, and offline evals with regression fixtures.

Hi Product Hunt - I built Industry Research Framework because long AI research tasks fail in repeatable ways: agents start before scope is clear, mix facts with judgment, leak process notes, or declare victory after a thin draft. This is a lightweight, open-source protocol for Codex, Claude, Gemini CLI, Cursor, OpenClaw, Hermes, and other agents. It gives the agent a brief gate, state files, source/claim/uncertainty registries, staged drafting, review loops, and hard stops for empty collection, evidence drift, false completion, depth collapse, and process leakage. It also ships eval cases, a source pack, rubric, gotchas, taste anchors, and regression fixtures so changes can be checked. I would love feedback from people using agents for market maps, investment memos, competitive analysis, or research writing: where should the protocol be stricter, lighter, or easier to adopt?

Love that this is open source and actually cares about source/claim discipline, not just vibes. The hard stops between stages feel like the kind of structure I always end up hacking together myself.

the brief gates and gotchas sections are a really thoughtful touch, shows the team has actually felt the pain of agent research going off the rails

Love that this is open source and actually cares about source/claim discipline, not just vibes. The hard stops between stages feel like the kind of structure I always end up hacking together myself.

the brief gates and gotchas sections are a really thoughtful touch, shows the team has actually felt the pain of agent research going off the rails

Industry Research Framework

Source-backed research protocol for AI agents

Source-backed research protocol for AI agents