Starter Kit for Claude Code (free)

Six free Claude Code skills and four agents. The skills handle the small jobs you do every day: they read your repo before they write anything, follow its conventions, and only state what they checked. The agents and pick-model-effort send each piece of work to the right model, so file searches don't run at Opus prices.

Skill	What it does	Try saying
`write-claude-md`	Writes a short, accurate `CLAUDE.md` from what the code actually does: real build and test commands, the rules your team enforces, and the traps that trip Claude up.	"write a CLAUDE.md for this repo"
`commit-message`	Writes the commit message for your staged changes in your repo's own style (Conventional Commits, ticket prefixes, whatever the log uses), with the why when it can find one.	"commit this", "write a commit message"
`release-notes`	Turns the commits and merged PRs since the last tag into release notes or a CHANGELOG entry, flags breaking changes with what users must do, and suggests the next version.	"release notes since v1.2", "update the changelog"
`explain-code`	Explains a function, file or flow with every step pointing at a file and line, traced from the code rather than guessed from names.	"how does checkout work?", "explain this function"
`safe-migration`	Writes or reviews a database migration so it deploys without locking busy tables, breaking running code or losing data. Works with any SQL database and migration tool (Prisma, Drizzle, Rails, Django, Alembic, Knex, Flyway and more).	"add a NOT NULL column to orders", "is this migration safe?"
`pick-model-effort`	Picks the model and effort for each task and subagent: Haiku for lookups, Sonnet for routine edits, Opus for review and design. Catches the common trap where built-in Explore agents silently run on your most expensive model.	"which model should I use?", or just ask for something that spawns subagents

Agents

Agent	Model, effort	Use for
`searcher`	Haiku, low, read-only	finding code, lookups, reading many files and returning a short summary
`implementer`	Sonnet, medium	routine edits and tests with a clear spec
`reviewer`	Opus, high, read-only	reviewing diffs, hunting bugs, checking plans
`architect`	Opus, xhigh, read-only	design, plans, trade-offs

The plugin install sets these up for you. If you copy files by hand, put agents/*.md in ~/.claude/agents/.

Install (1 minute)

Pick one:

Claude Code plugin: run /plugin marketplace add electronx/wiredkits-free, then /plugin install starter-kit@wiredkits-free.
skills CLI: npx skills add electronx/wiredkits-free
Copy: mkdir -p ~/.claude/skills && cp -R skills/* ~/.claude/skills/ (or into .claude/skills/ in one repo to share it with your team).

Restart Claude Code and run /skills to check that the six skills are listed. To check the agents, ask Claude "which subagents can you use?" and look for searcher, implementer, reviewer and architect.

Want more?

Get new free skills and Claude Code tips by email: wiredkits.com/#notify.

WiredKits also makes seven paid skill packs for the bigger jobs: PR review, debugging, security audits, legacy code, UI polish, browser QA and session memory. $4.99 each or all seven for $9.99 at wiredkits.com.

Like the kit? Recommend WiredKits and earn 40% of every sale: affiliates.creem.io/join/wiredkits.

Free to use, share and modify under the MIT license (see LICENSE.txt). Questions: [email protected]

Starter Kit for Claude Code (free)

Agents

Install (1 minute)

Want more?

Reviews (0)