chore: consolidate agent config via symlinks + add Claude Code lint/guard hooks#48
Merged
Merged
Conversation
Replace hand-mirrored duplicate agent config with single sources of truth: - AGENTS.md is now canonical (tool-neutral wording, AGENTS.md standard) - CLAUDE.md is a symlink -> AGENTS.md (Claude Code resolves it) - .agents/skills is a symlink -> ../.claude/skills - Drop the tool-specific attribution line in updating-deps-auto The previous per-commit mirroring had already drifted (AGENTS.md was missing 4 rules pointers; a mirrored skill referenced a broken .Codex/ path). Symlinks give zero drift with one place to edit. .claude/rules/ was already shared; .claude/agents/ and settings stay Claude-specific. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Aligns pinned pre-commit hook versions with backend deps (ruff>=0.15.12, mypy>=2.1.0). Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
…ooks Adds .claude/hooks/ scripts wired through settings.json: PreToolUse blocks hand-edits to frontend/api/generated/; PostToolUse runs ultracite check on changed frontend files (report-only) and ruff --fix + format on changed backend src/tests Python (mirrors .pre-commit-config.yaml; mypy left to pre-commit/CI). Codex hook sync intentionally skipped. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Pre-PR code review flagged that block-generated-edits.py returned 0 (allow) on any falsy file_path. For a security guard, fail open on an unexpected payload shape is the wrong default. Now: accept file_path or notebook_path, and if neither is present, block (exit 2) with a clear message. Lint hooks keep failing open by design (best-effort formatters). Unparseable stdin still returns 0 (no tool context). Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Generic boilerplate skill bulk-added in 5d9d900; not project-aware, redundant with IDE/tree, and emits an untracked codebase-map.html artifact on each run. No code or docs reference it. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
…oded list Phase 8 enumerated 6 fixed paths to `git add`, which drifts as the project evolves and implies docs must change every run. Replace with a git-status-driven flow: show the diff before the approval gate, skip the commit when nothing changed (no empty-commit error), then `git add -A`. The run is on its own update-deps/<date> branch with no auto-fixes, so committing everything that changed is correct and safe. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
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
1 participant
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.
Summary
Consolidates duplicated agent config to single sources of truth and adds Claude Code automation hooks. Four commits:
12430fc—AGENTS.mdis now the canonical agent-instructions file (tool-neutral, AGENTS.md standard).CLAUDE.md→ symlink →AGENTS.md;.agents/skills→ symlink →../.claude/skills. Eliminates the hand-mirroring that had already drifted (AGENTS.md was missing 4 rules pointers; a mirrored skill referenced a broken.Codex/path)..claude/rules/was already shared;.claude/agents/and settings stay Claude-specific by design.484bde9— pre-commit bump: ruffv0.14.10→v0.15.13, mypyv1.19.1→v2.1.0(matchesbackend/pyproject.tomlfloors).fdb376f— new Claude Code hooks:backend-ruff.py+frontend-lint.py(PostToolUse formatters),block-generated-edits.py(PreToolUse guard forfrontend/api/generated/), wired in.claude/settings.json. Claude-Code-only by design — git pre-commit covers backend lint; these fill the frontend/generated-guard gap for agent sessions.d0b33b5— review fix: the generated-file guard now fails closed (blocks) when it can't determine a target path, instead of silently allowing the edit.Review
Reviewed pre-PR via the
code-reviewersubagent → fix-and-ship. Symlink consolidation verified (mode120000, resolves correctly, no stale.Codex/refs). The one substantive finding (guard fail-open) is addressed ind0b33b5; the reviewer's MultiEdit-specific premise was inaccurate (MultiEdit does carryfile_path), but the fail-closed hardening is the correct defense-in-depth regardless. Guard behavior verified across 5 payload shapes (generated/normal/notebook/no-path/unparseable).Caveat
Symlinks are git-tracked as mode
120000; they degrade to plain text files on a Windows checkout withoutcore.symlinks. Accepted — macOS target environment.🤖 Generated with Claude Code