docs(agents): policy updates — uv tooling, GPU caveat, Conventional Commits, no AI attribution#110
Merged
Conversation
The checklist commands didn't match CONTRIBUTING.md's uv-managed invocations, implied a green local pytest validates kernel changes (CUDA tests auto-skip without a GPU), and said nothing about keeping the prose docs in sync with the code — the source of most of the doc drift this PR fixes.
Half the file was explanation around rules agents already follow once stated. Keep one line per rule, add the H100 bench-validation and GPU-CI-label protocol that PR #109 proved necessary, require Conventional Commits, and ban AI attribution in commits and PRs.
Claude Code reads only CLAUDE.md; the @AGENTS.md import is the documented way to apply the shared agent rules there (symlinks need admin rights on Windows and leave no room for Claude-specific notes).
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Context
AGENTS.md's checklist used bare
ruff/pytestcommands instead of CONTRIBUTING.md'suv runtooling, and nothing covered the GPU-validation gap (CUDA tests auto-skip locally, so local green doesn't validate kernel changes), the pre-PR docs sweep requirement, or the H100 benchmark-validation and GPU-CI protocol that #109 proved necessary. Two policies were also implicit but unwritten: Conventional Commits format and no AI attribution in commits or PR text.Separated from #108 (doc accuracy sweep). These are policy additions, not doc-drift corrections.
Changes
Click here to expand
AGENTS.md— commands updated touv runinvocations; pre-PR docs sweep step added (checkhow-it-works.html,design.md, README tables for invalidated claims); H100 bench-validation andrun-gpu-testsGPU-CI protocol added; Conventional Commits (type(scope): subject) required; AI attribution banned in commits, PR descriptions, and PR comments. Condensed from 108 to 33 lines.CLAUDE.md— new file with a one-line@AGENTS.mdimport. Claude Code reads onlyCLAUDE.md; the@import applies the shared agent rules without a symlink (cross-platform, the documented pattern)..gitignore— ignores.claude/worktrees/(ephemeral agent worktrees).Validation
No code changes; no test run required.