Senior Dev Skills for AI Coding Agents
Coding agent skills to think like a Senior Engineer.
By TheSeniorDev · MIT-licensed · Contributions welcome
Install
npx skills@latest add the-senior-dev/senior-dev-skills
Or with Claude Code's native plugin marketplace:
/plugin marketplace add the-senior-dev/senior-dev-skills
/plugin install senior-dev-skills@senior-dev-skills
Restart Claude Code (or open a new session) to pick up the new skills.
Need Claude Code? Get it at claude.com/code.
How to use in Claude Code
Invoke a skill by name, with an optional argument:
/react-senior-code-review src/features/checkout # review a feature
/react-senior-interview state # quiz yourself on state & data flow
react-senior-code-review takes a path to the feature you want reviewed; react-senior-interview takes an optional topic (or none, for a broad interview).
What's inside
These skills are like having a Senior Engineer by your side, pointing out things you might have missed. They leverage Software Fundamentals, Design Patterns, and Senior Mental Models — together with the power of Claude Code — to push your code toward production readiness, prep you with real interview questions, and ground every call in deep domain knowledge and fundamentals. Keep in mind they are:
- Highly opinionated. Biased toward what holds up in production — the wisdom a senior reviewer brings, not a neutral checklist.
- Anti-
useEffect. Reach for derived state first, then event handlers, then render-time computation. Effects are the last resort.
- Server state ≠ client state. Server data lives in a query layer (TanStack Query, SWR, route loaders, RSC) — never in
useState+useEffect, never in Zustand.
- Push state down. State lives as close to its consumer as possible.
- Colocate, then share. Code used by one feature stays in that feature; promote it to
shared/ only when a second feature actually needs it.
- Name the pattern. Custom hook? Compound components? HOC? Render prop? State machine? Optimistic update? If a known pattern fits the fix, it gets named.
Both skills run on the same rules catalog — one file per dimension, plus Rules of Hooks, the patterns playbook, the severity rubric, and curated docs. The canonical copy lives in skills/react-senior-code-review/references/ (indexed by principles.md); react-senior-interview carries a synced copy so each skill installs and runs independently. Edit the canonical files, then run scripts/sync-references.sh.
Roadmap
Each skill ships an authoritative knowledge catalog (a references/ folder indexed by principles.md) and a thin SKILL.md that runs the workflow. Planned additions in the same monorepo:
node-senior-review / node-interview
typescript-senior-review
system-design-interview
You can open an issue with the area you want next.
Contributing
The principles in skills/react-senior-code-review/references/ are the core of both skills. Pull requests are welcome — especially:
- Sharper rule statements (less verbose).
- New patterns (with a clear "when it makes sense / when not").
- Better calibration for interview questions.
If you're adding a new skill, mirror the layout: SKILL.md is the workflow, a references/ folder (indexed by principles.md) is the knowledge.
License
MIT — use freely, fork freely, ship freely.