verification-before-completion_skill

Overview

This skill implements a five-gate verification workflow to validate work before marking it complete. It enforces functional, technical, testing, security, and deployment quality gates so issues are caught early and releases remain reliable. The skill provides language-specific scripts and CI hooks to automate checks and produce actionable results.

How this skill works

Run the provided verification runner (bash script or Python/TypeScript gate scripts) to execute a sequence of checks: formatting, linting, type checks, tests with coverage, security scans, and build verification. Each gate records pass/fail status and emits clear output so failures can be triaged. Integrate the gates into pre-commit hooks and CI pipelines to prevent incomplete or low-quality changes from being merged or deployed.

When to use it

• Before marking a feature or bugfix as done or requesting a review
• Prior to merging a pull request that affects production behavior
• Before deploying to staging or production environments
• When enforcing team quality gates or release checklists
• When you need automated, reproducible verification across languages

Best practices

• Run quality gates locally via pre-commit hooks to catch issues early
• Keep acceptance criteria testable and write tests alongside changes
• Treat gate failures as tickets: fix, re-run, and document the resolution
• Tune coverage thresholds and security scan levels to match risk
• Integrate gate output into CI results and PR checks for visibility

Example use cases

• A Python feature branch runs python-quality-gates.py to ensure formatting, mypy, pytest coverage, and bandit checks pass before a PR
• A TypeScript change triggers the CI quality-gates workflow to run ESLint, prettier, tsc, tests, and npm audit on every push
• A release engineer runs the comprehensive verify-completion.sh script before a production deploy to validate migrations, build, and monitoring hooks
• Developers add the pre-commit script to prevent commits that fail formatting, linting, or unit tests