dev-skills

dev-skills

A collection of general software-development agent skills. Installable via either the Vercel Labs skills CLI or Claude Code's built-in plugin system.

Install

Via Claude Code (recommended for Claude Code users)

# Add this repo as a plugin marketplace, then install
/plugin marketplace add SebastianElvis/dev-skills
/plugin install dev-skills@dev-skills

After installing, run /reload-plugins to activate. Skills are namespaced under the plugin name, e.g. invoke pr-gen as /dev-skills:pr-gen. You can also browse and manage installed plugins via the /plugin interactive menu.

Via the skills CLI (works with any agent)

# All skills in this repo
npx skills add SebastianElvis/dev-skills

# Just one skill
npx skills add SebastianElvis/dev-skills --skill pr-gen

# Global install (all projects)
npx skills add SebastianElvis/dev-skills -g

Works with Claude Code, Codex, Cursor, OpenCode, and every other agent the skills CLI supports.

Skills

Skill	What it does
pr-gen	Generate or update a GitHub PR title and description from the actual code changes on the current branch.

Layout

.claude-plugin/marketplace.json   # Marketplace catalog (lists this plugin)
.claude-plugin/plugin.json        # Claude Code plugin manifest
skills/<name>/SKILL.md            # Each skill (standard location for both ecosystems)

The repo doubles as both a marketplace (catalog at .claude-plugin/marketplace.json) and the plugin itself (manifest at .claude-plugin/plugin.json, source "./"), so a single /plugin marketplace add registers the catalog and /plugin install pulls the plugin from the same repo.

Adding a new skill

cd skills && npx skills init <skill-name>

Then edit skills/<skill-name>/SKILL.md and add it to the table above.

Evaluating skills

The repo ships an eval harness under evals/, modeled on Anthropic's Demystifying Evals for AI Agents. Both the runner and the LLM judge shell out to claude -p (no API key required).

# Does the skill trigger on the right prompts and skip the wrong ones?
python3 evals/run.py triggering --skill pr-gen --trials 2 --parallel 6

# Given it triggered, is the output good? Runs against fresh tmp git repos.
python3 evals/run.py functional --skill pr-gen

See evals/README.md for layout, judge design, and how to add cases for a new skill.

License

MIT