zilianglab/consequence-gate

A governance layer between an agent loop and its tools: resolves each tool call to an autonomy tier from declared consequence metadata (not model confidence), with an append-only audit log and a context-aware human-override loop.

1 best points · 0 comments · 1 Hacker News post · 1 GitHub stars · View on GitHub

Hacker News discussions

• Show HN: Consequence Gate – agent governance based on what being wrong costs (1 points, 0 comments, by ziliangh, 2026-08-28)