regression-guard · v1.0.0 · 2026-05-23 · sha256 2c8f77d5a0fdbd82
regression-guard v1.0.0A
Immutable. This exact content never changes and is served at /api/v1/blob/2c8f77d5a0fdbd82.
No tool to install first
claude plugin marketplace add https://markdownregistry.com/mp/a/art_awi7k6dfrmqwpqnf/2c8f77d5a0fdbd828ec948b98893e9ee81a86a4833a6d1aefeb9077e8c1a7044.json
Then /plugin install regression-guard@markdownregistry-awi7k6dfrmqwpqnf in Claude Code. It installs this exact version, the one graded A here, pinned to the commit it was graded at, and leaves it there.
--- name: regression-guard description: "Post-fix test-coverage analysis. Reviews a just-landed fix, generates targeted unit and regression tests, audits how the bug could have been caught earlier. Use after a fix is merged, when asked to 'add regression tests', 'pay off test debt', 'guard against regression', or 'ensure this doesn't happen again'." version: 1.0.0 user_invocable: true --- # Regression Guard Post-fix test-debt payoff loop. Takes the git diff of a recent fix as input and generates targeted tests to prevent the bug from returning. ## Usage ``` /regression-guard [commit-range] ``` If no commit range is specified, uses `HEAD~3..HEAD`. ## Step 1 — Review What Was Fixed Build a structured summary from git: ```bash git log --oneline -10 git diff HEAD~3..HEAD git log -1 --format=%B ``` Produce: - **What broke** — the failure mode - **What changed** — specific code paths - **Files and functions touched** - **Issue/PR reference** ## Step 2 — Generate Unit Tests Write the smallest set of unit tests that lock in the correct behavior: ``` tests/unit/<area>/<fix-slug>.test.<ext> ``` Rules: - Each test must be red when reverted against the pre-fix state - Cover the happy path AND the corrected edge cases - State which assertion would be red if reverted ## Step 3 — Generate Regression Tests Write regression tests that construct the exact pre-fix state: ``` tests/regression/<area>/<fix-slug>-regression.test.<ext> ``` Rules: - Include a top-of-file comment identifying the pre-fix state and symptom - The test must pass the full check suite with no manual edits ## Step 4 — Audit: Could This Have Been Caught Earlier? Answer honestly. Pick the most relevant gap category: - **Missing boundary test** — contract assertion at a module seam - **Missing negative test** — error path never exercised - **Non-hermetic test harness** — shared state leaked between tests - **Missing integration coverage** — only reproduced under production shape - **Contract drift** — upstream contract widened/narrowed without downstream rebase For each gap, write a concrete, actionable suggestion. No generic "add more tests." ## Step 5 — Recommend Integration Testing Additions Name: 1. The specific test command to extend 2. The specific file to add a test into (or create) 3. A concrete test stub that would have caught the bug ## Output Report Write to `docs/investigations/<date>-regression-guard-<fix-slug>.md`: 1. Fix summary 2. Generated unit tests (file paths + descriptions) 3. Generated regression tests (file paths + pre-fix state each reconstructs) 4. "Could this have been caught earlier?" diagnosis 5. Integration testing recommendations 6. Baseline comparison table ## Ship as a Draft PR 1. Create a new branch 2. Commit the generated test files and output report 3. Open a draft PR against main 4. Never push directly to main