gear_skill

Overview

This skill is the DevOps mechanic for dependency management, CI/CD optimization, Docker configuration, and observability fixes. It focuses on small, high-impact maintenance tasks: a single build error, one safe dependency update, or an observability improvement per session. The goal is reproducible builds, fast feedback loops, and minimal, reviewable changes.

How this skill works

Gear inspects build logs, lockfiles, CI workflows, Dockerfiles, and observability configs to identify the highest-value maintenance action. It applies conservative fixes (patch/minor dependency updates, cache tweaks, Docker layer reordering, or small CI refactors), runs the build/tests, and generates a concise PR with verification notes and risk level. It logs the change and triggers handoffs when larger migrations or security reviews are required.

When to use it

• A single CI step fails or cache misses frequently
• Builds are slow due to Docker or dependency issues
• A dependency audit shows actionable vulnerabilities or rot
• You need a small linter/formatter or git-hook config fix
• Observability is missing health checks, logs, or a basic Sentry/Prometheus hook

Best practices

• Respect SemVer: prefer patch/minor updates and ask before majors
• Keep changes small (<50 lines) and verifiable locally or in CI
• Regenerate and commit lockfiles; use --frozen-lockfile in CI when possible
• Use composite/reusable GitHub Actions and cache layers for DRY pipelines
• Log each maintenance action in the project agent journal and add a short activity entry

Example use cases

• Fix a flaky build step by restoring a cache key and updating the workflow to use BuildKit
• Resolve lockfile conflicts and run pnpm dedupe + audit fix, then verify the app starts
• Optimize a Dockerfile with multi-stage build and reordered COPY for smaller image layers
• Add ESLint/Prettier configs and Husky hooks, run lint in CI to catch regressions
• Add a basic /health endpoint and configure Prometheus metrics and Sentry for error reporting