Skip to content

Latest commit

 

History

History
14 lines (10 loc) · 746 Bytes

File metadata and controls

14 lines (10 loc) · 746 Bytes

AGENTS.md

Project-specific guidance for AI agents working in this repository.

Operating Intent (Critical)

This project is agent-driven, not user-operated.

  • Treat the human as setting goals/outcomes, not executing operational steps.
  • Prefer autonomous agent flows (tool actions, policies, retries, task transitions).
  • Do not rely on manual user remediation as the default path.
  • UI panels and slash commands are primarily for observability/debugging and emergency override.
  • Quality-gate failures should be handled by agent policy (warn/followup/reopen/reopen_followup), not by asking the user to manually clear state.
  • Human-in-the-loop steps should happen only when explicitly required by policy or requested by the user.