Strict pre-merge quality check suite for production reliability.

Install

mkdir -p .claude/skills/quality-gate && curl -L -o skill.zip "https://agentskills.codes/api/skills/download/10222" && unzip -o skill.zip -d .claude/skills/quality-gate && rm skill.zip

Installs to .claude/skills/quality-gate

Activation

This is the description your AI agent reads to decide when to run this skill — the better it matches your request, the more reliably it fires.

Mandatory pre-merge quality checks for production code — security scan, packaging test, API consistency, no silent failures. Prevents the issues that caused 4 emergency releases in v0.21.x.
189 charsno explicit “when” trigger
Intermediate

Key capabilities

  • →Run security scans
  • →Validate API consistency
  • →Verify build integrity
  • →Check architecture boundaries
  • →Enforce coding standards

How it works

It executes a series of automated checks at different levels to verify code stability before merging.

Inputs & outputs

You give it
code changes
You get back
quality gate report

When to use quality-gate

  • →Run pre-merge quality gate
  • →Validate API consistency
  • →Verify build integrity
  • →Prevent regression in production releases

About this skill

Quality Gate

Use the repository's canonical gates from the workspace root. Do not recreate them with ad hoc greps, repeated test suites, or persistent cd commands.

When to Use

  • Before committing a production change
  • Before a PR is accepted or merged
  • At implementation closeout
  • Before a release

Verification Ladder

1. Implement the bounded scope first

Complete the scoped code, tests, documentation, evidence, and other intended writes before the routine verification sequence. Do not treat every saved edit as a gate boundary.

2. While editing: diagnose only when needed

When a current question or failure requires evidence, run only the smallest existing check that exercises the changed main process. Do not rerun a check merely because another edit was made. Examples:

./scripts/python_runtime.sh -m pytest Python/tests/path/to/test_file.py -q
./scripts/python_runtime.sh -m pytest fastapi_app/tests/path/to/test_file.py -q
npm --prefix react_app run lint
./scripts/python_runtime.sh scripts/check_architecture_boundaries.py

Choose commands from the affected component's skill or existing project automation. Do not add tests during a review-only task.

3. After content freeze: format and focused evidence together

After all intended versioned writes are complete, run the affected focused tests, benchmarks, and architecture/import checks as one consolidated selection. First run ./run.sh format --write; it selects only changed Python, FastAPI, and C# paths and fails if formatter bytes escape that scope. Validate live repository context before this selection when routing or repository structure is affected; this check is read-only.

4. Accepted candidate: integrity once

./run.sh check --candidate-integrity

Run this read-only hosted-equivalent file check only after the independent audit accepts the immutable candidate. If it fails, the candidate is not immutable: record INTEGRITY_REJECTED, return to the writer state, repair the root cause, and create the one allowed repair candidate. A failure on that repaired candidate enters REPLAN. Do not run check --quick as a ritual before every commit.

5. PR acceptance: required CI

The GitHub check named PR Gate is the authoritative merge gate. Inspect the check for the current commit. Do not rerun an equivalent local suite merely because CI already passed it. A failing or stale check blocks acceptance; bypass flags and admin merges are forbidden. Record a failed hosted run as HOSTED_REJECTED before repairing its confirmed root cause. The replacement candidate must repeat exact-head audit, integrity, pre-push closeout, and one hosted verdict; retain the failed run in closeout metrics.

6. Cumulative implementation closeout: full gate once

./run.sh check

Run the full gate once after all intended packets in the milestone are integrated. Run it earlier only when repository-wide risk makes that necessary. After a failure, rerun only the failed check while repairing it; repeat the full gate only when the fix can affect other categories or to establish the final green result.

7. Release only

For an actual release candidate, use the release skill and its canonical preflight:

./run.sh release preflight <version>

Do not run packaging, Docker, UAT, or release checks for an ordinary review or commit.

Essential-Only Review Rule

For each possible finding ask: Would fixing it change the outcome of the main process in scope? Report and block only when the answer is yes and the defect is confirmed by evidence. Ignore comments, adjacent improvements, coverage gaps, generic hardening, and security or concurrency observations that do not change that outcome. Do not add tests during review.

Report Format

## Quality Gate Report

| Check | Status | Details |
|-------|--------|---------|
| [check name] | ✅/❌/⚠️ | [details] |

**Gate:** PASS / FAIL
**Current commit:** [hash]
**Blockers:** [only confirmed main-process failures]
**Not rerun:** [checks already covered by green CI, if any]

Who Uses This

  • @reviewer — evaluates main-process findings and PR Gate evidence
  • @ops — runs the safe Git/CI workflow and release preflight
  • @tester — runs narrow behavior checks and the canonical closeout gate
  • @governance — keeps the gate definitions truthful and non-duplicative

When not to use it

  • →When code is not for production
  • →When skipping pre-merge checks

Limitations

  • →Mandatory pre-merge

How it compares

It provides mandatory, tiered quality checks specifically intended to prevent historical failure patterns.

Compared to similar skills

quality-gate side by side with the closest alternatives in the catalog.

SkillInstallsUpdatedSafetyDifficulty
quality-gate (this skill)05moReviewIntermediate
e2e-testing-patterns84moNo flagsIntermediate
testing-workflow1611moReviewIntermediate
perf-lighthouse137moReviewIntermediate

Try saying

Example prompts that trigger this skill in your AI assistant.

You might also like

Search skills

Search the agent skills registry