Engineering Skills

Languages: English | 中文

Build Claude Code Python version License: MIT

A Claude Code plugin — 47 engineering skills organized by the software development lifecycle. Shared engineering principles are injected at every session start, so the discipline is ambient — not something you have to remember to load. PM and UIUX skills additionally layer product-principles.md / design-principles.md as domain discipline.

Doc-producing skills split output into two layers: project-level (docs/PRD.md etc., concise, on main) and version-level (docs/vX.Y/prd.md etc., detailed, on version branches). See docs/skill-outputs.md for the full matrix.

Install

Claude Code (primary):

/plugin marketplace add https://github.com/int2t05/engineering-skills
/plugin install engineering-skills@int2t05

Other agents (Codex, Cursor, Cline, Continue, OpenCode, Windsurf…): skill content is plain markdown. See .agents/install-block.md for the canonical install commands.

How to use

  • Auto-trigger: skills activate when a task matches their description (English or Chinese — e.g. "技术选型", "性能优化", "深度检索").
  • Explicit call: type /skill-name (/tdd, /code-review, /debugging).
  • Not sure which skill? /using-skills routes the task to a phase.
  • Planning: uses the harness's built-in plan mode, not a custom skill.

Structure

engineering-skills/
├── .claude-plugin/          # plugin.json (manifest) + marketplace.json
├── .github/
│   ├── workflows/validate.yml   # CI: schema + adapter-sync + invocation-sync
│   ├── ISSUE_TEMPLATE/          # bug-report.md
│   └── PULL_REQUEST_TEMPLATE.md
├── skills/                   # 47 skills, 9 SDLC phases + meta
│   ├── meta/using-skills/       # router
│   ├── meta/skill-authoring/    # author + eval skills
│   ├── 01-product/ … 09-operate/
│   └── <skill>/SKILL.md + references/ + agents/openai.yaml
├── references/               # engineering / product / design principles (shared)
├── scripts/                  # validate-skills.sh, gen-agents-yaml.py, run-eval, check-routing-overlap.py
├── docs/                     # skill-outputs.md, workflow-prompts.md
├── AGENTS.md                 # universal entry for non-Claude agents
└── LICENSE                   # MIT

Catalog — 47 skills by phase

Phase Skills
meta using-skills, skill-authoring
01-product brainstorm, spec, oss-strategy
02-research research (general/market/tech-selection modes)
03-design architecture, api-design, frontend-design, schema-design, prompt-engineering, domain-modeling, prototype, codebase-design, image-to-code, imagegen, design-research
04-develop implement, multi-agent-orchestration, breakdown, context-engineering, i18n, auth-implementation, error-handling
05-tune cost-optimization, performance, simplify, refactoring
06-test tdd, test-generation, api-testing, e2e-testing, load-testing
07-verify a11y-review, code-review, debugging, security-review, linting
08-ship shipping, git-workflow, ci-cd, deprecation-migration, oss-polish
09-operate observability, documentation-audit, incident-response, handoff

Triggers + routing: AGENTS.md · Product matrix: docs/skill-outputs.md

Validate the collection: bash scripts/validate-skills.sh

Reporting Issues

Bugs and suggestions: github.com/int2t05/engineering-skills/issues

License

MIT — see LICENSE.

Provenance

Original skill concepts adapted from addyosmani/agent-skills and mattpocock/skills. This pack is a ground-up rewrite fusing those concepts with systematic PM/UIUX principles.