Engineering Agent Skills

Make your coding agent a stronger engineering partner for code reviews, implementation planning, and production-grade system design.

This package installs a ready-to-use agent skill that helps your agent reason like a practical senior engineer: clear tradeoffs, focused changes, fewer hidden risks, and better engineering communication.

Why Use It

| Need | How the skill helps | | --- | --- | | Better code reviews | Finds correctness gaps, missing tests, maintainability issues, and production risks. | | Cleaner implementation plans | Keeps changes focused, testable, and aligned with existing module boundaries. | | Stronger system design | Guides architecture tradeoffs for feeds, search, queues, caching, workflows, payments, LLM serving, and more. | | Sharper engineering writing | Improves commit messages, API docs, release notes, ADRs, and PR feedback. |

What You Get

• Practical coding principles for simple, maintainable changes without unnecessary abstraction.
• Low-level design guidance for module boundaries, state ownership, APIs, concurrency, resource safety, and testable implementation plans.
• Staff-level system design playbooks for common production systems, including feeds, search, queues, caching, rate limiting, real-time messaging, workflows, monitoring, large-file handling, payments, LLM serving, and high-contention inventory.
• Commit message guidance so every change has clear scope, behavior, and validation.
• Review checklists for architecture decisions, API docs, PRs, release notes, missing tests, and production readiness.

Quick Start

Run the installer in the project where you want your agent to use the skill:

npx --yes engineering-agent-skills@latest install

This copies the bundled skill into the directory where you ran the command:

.agents/skills/engineering-agent-skills

This is the cleanest install path because it does not add a project dependency, node_modules, or lockfile changes.

If you want to pin the skill package as a project dependency instead, use:

npm install engineering-agent-skills

The npm install path uses the same default target: .agents/skills with an s.

Use It

Ask your agent to use the skill by name:

Use engineering-agent-skills to review this architecture proposal.

Try it for everyday engineering work:

Use engineering-agent-skills before implementing this feature.
Use engineering-agent-skills to review these API docs before release.
Use engineering-agent-skills to write a commit message for this diff.
Use engineering-agent-skills to design a hybrid follower timeline.
Use engineering-agent-skills to review state ownership and concurrency risks.

Popular Workflows

| Workflow | Example prompt | | --- | --- | | Code review | Use engineering-agent-skills to review this PR for production risks. | | Implementation planning | Use engineering-agent-skills to plan the smallest safe implementation for this feature. | | System design | Use engineering-agent-skills to design a rate limiter for this API. | | API documentation | Use engineering-agent-skills to review these endpoint docs before release. | | Commit messages | Use engineering-agent-skills to write a commit message for this diff. |

Command Reference

Install into the default project skills directory:

npx --yes engineering-agent-skills@latest install

Install for Claude Code project skills:

npx --yes engineering-agent-skills@latest install --agent claude-code

Install for Codex user-global skills:

npx --yes engineering-agent-skills@latest install --target ~/.codex/skills

Install into a custom skills directory:

npx --yes engineering-agent-skills@latest install --target .agents/skills

List bundled skills in this npm package:

npx --yes engineering-agent-skills@latest list

list does not inspect your local .agents/skills, .claude/skills, or ~/.codex/skills directories. It only prints the skill names bundled in the package.

Best Fit

Use this skill when you want your agent to slow down just enough to catch engineering risk before it ships: unclear ownership, brittle abstractions, missing tests, weak API contracts, unsafe concurrency, vague commit messages, or architecture decisions that need sharper tradeoff analysis.