Andreas Fehlner
cb9eb1006d
## Summary Closes a few gaps found while auditing this repo against a draft "agent control" rubric (agents.md/claude.md, contribution guidelines, lint enforcement, security.md cross-linking): - `CLAUDE.md` now tells agents to review `CONTRIBUTING.md` before committing, pushing, filing an issue, or opening a PR, and to follow `SECURITY.md`'s disclosure process for anything security-related instead of filing a public issue. - `CLAUDE.md` now states explicitly that `lintrunner` must pass with no errors before a coding task is considered complete (previously it only documented how to invoke it). - Adds `AGENTS.md` as a symlink to `CLAUDE.md`, so tooling that looks for the `AGENTS.md` convention (e.g. Codex-style agents) picks up the same guidance without duplicating content. ## Note on tooling This repo has a `pixi`-based dev workflow (`CONTRIBUTING.md`'s "Pixi-based development" section). I considered wiring the lint instructions in `CLAUDE.md` through `pixi run lintrunner` instead of a bare `lintrunner` invocation, but held off — not everyone's agent/dev setup has `pixi` installed, and `lintrunner` works standalone. Flagging here in case maintainers want to standardize agent-facing commands on `pixi` going forward; happy to follow up if so. ## Test plan - [x] Verified `AGENTS.md` is committed as an actual symlink (`git ls-files -s AGENTS.md` shows mode `120000`) pointing at `CLAUDE.md`. - [ ] CI --------- Signed-off-by: Andreas Fehlner <fehlner@arcor.de>
Symbolic link
9 B
Symbolic link
9 B
CLAUDE.md