A curated set of skills for each stage of development — propose, spec, design, plan, implement, ship.
npx claudepluginhub codagent-ai/agent-skillsA curated set of skills for each stage of development — propose, spec, design, plan, implement, ship.
Codagent-AI
Codagent is a plugin for Claude Code and Cursor that provides a set of focused skills for spec driven development — from evaluating an idea to implementing with subagents to shepherding a PR through CI. Inspired by obra/superpowers and OpenSpec.
Each skill is invoked with /codagent:<skill-name>.
init — Initializes Codagent in your project. Checks that the Agent Validator CLI is installed and configured. Safe to re-run.propose — Evaluates whether an idea is worth building. Researches the codebase and web, delivers a GO / GO WITH CAVEATS / NO-GO verdict, and writes proposal.md.spec — Drives interactive requirement discovery. Walks through each capability from the proposal, asking about behaviors, boundaries, error conditions, and edge cases. Produces spec files with testable WHEN/THEN scenarios.design — Creates a technical design through collaborative brainstorming. Proposes 2-3 approaches with trade-offs, presents the design incrementally for approval, and writes design.md.review-spec — Reviews design artifacts (proposal, specs, design, tasks) for internal consistency, cross-artifact alignment, and gaps. Reports findings with exact citations.plan-tasks — Creates a structured task breakdown from the proposal, design, and specs. Each task file is self-contained with everything a subagent needs to implement it.implement-change — Autonomous tech lead that implements a full change end-to-end. Dispatches one implement-and-validate subagent per task (sequentially, never in parallel), runs the gauntlet, archives the change, and finalizes the PR.implement-with-tdd — Enforces test-driven development: write a failing test, watch it fail, write minimal code to pass, refactor. No production code without a failing test first.implement-and-validate — Autonomous implementer that executes a single task end-to-end. Calls implement-with-tdd to build the code, performs self-review, runs the Agent Validator, and commits on success.push-pr — Commits changes, pushes to remote, and creates or updates a pull request. Runs the validator before committing if applicable.wait-ci — Polls CI status for the current branch's PR. Enriches failures with GitHub Actions logs, surfaces blocking reviews and unresolved PR comments.fix-pr — Fixes CI failures and review comments on the current branch's PR. Dispatches a fixer subagent, verifies with the validator, and pushes.finalize-pr — Orchestrates the full post-implementation loop: push PR, wait for CI, fix failures, repeat until green. Stops after 3 fix cycles or when the same failure persists.Codagent requires the Agent Validator CLI for automated quality verification:
npm install -g agent-validator
agent-validator init
claude plugin marketplace add Codagent-AI/agent-skills
claude plugin install codagent
cursor plugins install Codagent-AI/agent-skills
/codagent:init
claude plugin marketplace update codagent
claude plugin update codagent@codagent
Then run to get the latest skills:
/codagent:init
MIT
Official prompts.chat marketplace - AI prompts, skills, and tools for Claude Code
Open Design — local-first design app exposed to coding agents over MCP. Install once with your agent's plugin command and projects/files/skills are reachable through stdio.
Behavioral guidelines to reduce common LLM coding mistakes, derived from Andrej Karpathy's observations