Skip to content

hooks.md

Latest commit

 

History

History
37 lines (25 loc) · 1.74 KB

hooks.md

File metadata and controls

37 lines (25 loc) · 1.74 KB
Raw
alwaysApply true
description Enforces correct interaction with pre-commit, post-edit, and post-test hooks. Never bypass hooks; investigate failures instead.

Hook Rules

Format Hooks

  • DO auto-accept post-edit format hooks. They enforce consistent style automatically. Rationale: Format hooks maintain codebase consistency. Fighting them creates churn.

  • DON'T revert or undo formatting changes applied by hooks. Rationale: The hook output is the canonical style. Manual overrides cause style drift.

Pre-Commit Hooks

  • DON'T skip pre-commit hooks with --no-verify. Ever. Rationale: Pre-commit hooks catch real issues (build errors, lint failures, format violations). Bypassing them pushes broken code.

  • DO investigate and fix the root cause when a hook blocks a commit. Rationale: The hook is telling you something is wrong. Silencing it hides the problem.

Post-Test Analysis

  • DO pipe test output through hooks/post-test-analyze.sh when running test workflows (dotnet test 2>&1 | bash hooks/post-test-analyze.sh), and act on its summary. Rationale: The structured summary surfaces failures and next steps that raw test output buries.

Hook Infrastructure

  • DON'T interfere with hook configuration. Claude Code hooks (pre-bash-guard, post-edit-format, post-scaffold-restore) run automatically via hooks/hooks.json; pre-commit scripts run via git. See hooks/README.md for the full map. Rationale: Hooks are configured intentionally. Ad-hoc changes break the automated workflow.

  • DO wait for post-scaffold-restore to complete after .csproj changes before building. Rationale: NuGet restore must finish before the build can resolve dependencies. Building too early produces false errors.