Skip to content

Latest commit

 

History

History
31 lines (24 loc) · 1.43 KB

File metadata and controls

31 lines (24 loc) · 1.43 KB

AGENTS.md — LoopCheck

Read CLAUDE.md first, and follow it.

CLAUDE.md is the single source of truth for this repository: what LoopCheck is, the hard constraints, the locked tech stack, the data model, the seed library, the conventions, the security posture, the docs-as-code definition of done, and the execution/roadmap-maintenance workflows.

This file exists only because some agents look for AGENTS.md by name. It deliberately does not restate those rules — a second copy is a second thing to drift, and this file had already drifted out of date before it was reduced to a pointer. Everything that applies to any agent working here lives in CLAUDE.md.

The only agent-specific note

Git authorship applies to every agent, not just Claude. Commits are authored as the maintainer, with no AI-attribution trailer of any kind — no Co-Authored-By, no "generated by" line — regardless of which assistant made the change. This is also enforced by .claude/settings.json (that is the real path; there is no .Codex/ directory in this repo).

Where to start

  1. CLAUDE.md — constraints and conventions.
  2. docs/ROADMAP.md — the milestone sequence.
  3. docs/DECISIONS.md — what was decided, and what was deliberately not built.
  4. docs/tasks/ — pick the lowest-numbered task whose Status: is not DONE, and do exactly that task.