Skip to content

Cut agent instructions down to what the code can't tell an agent - #3

Draft
Moritz Mazetti (Radiergummi) wants to merge 1 commit into
mainfrom
prune-agent-docs
Draft

Moritz Mazetti (Radiergummi) wants to merge 1 commit into
mainfrom
prune-agent-docs

Conversation

@Radiergummi

Copy link
Copy Markdown
Member

Recent research suggests that large agent instruction files make coding agents perform worse. This PR keeps only what an agent cannot learn from the code, config or README, and turns rules into checks where the existing tooling allows it.

Summary

  • CLAUDE.md: 53 → 6 lines.

Check that replaces a prose rule

  • .claude/settings.json: a PreToolUse hook blocks agent edits to the generated .matchory/* and .editorconfig and names the sync command. It needs jq; CI's style check still catches hand edits.\n\nAll five CI checks pass.\n\n## Please check\n- Dropped the note that the format-on-save hook does nothing without jq.

🤖 Generated with Claude Code

Move the 'never hand-edit .matchory/ or .editorconfig' rule into a
PreToolUse hook that blocks those edits and names the sync command.
Drop the stack list, commands and rename guide (all in the README and
manifests), the tests/__init__.py note (ruff INP001 already fails
without it) and the PostToolUse hook description.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant