Agent Config

My configuration for working with AI coding agents — currently Claude Code and Pi. Pairs well with my agent-tools.

This repo is opinionated. It provides a structured development workflow, security-first hooks, and reusable skills that turn a general-purpose coding agent into a reliable development partner. Use it as-is, fork it, or cherry-pick the parts that fit your setup.

What's Included

Claude Code → ~/.claude/

• Structured development workflow — /brainstorming → /writing-plans → /executing-plans → /verifying-work → /completing-work, a pipeline of skills (adapted from superpowers) that turns ideas into pull requests with subagent-isolated implementation and parallel reviewers
• Reference skills — TDD discipline, PR review, browser automation (Playwright), Jira ticket creation, frontend design, incident troubleshooting, agent engineering, and more
• Security and quality hooks — Pre-commit secret scanning with gitleaks; auto-formatting on every write via Prettier, gofmt, rustfmt, or shfmt
• Sandbox mode — Locked-down config for headless or remote environments that redirects gh and remote git to MCP tools
• Custom status line — Powerline-style display showing model, branch, context window usage, and session rate-limit usage

Pi → ~/.pi/agent/

• Workflow-aware agent setup — Pi-specific AGENTS.md, model/settings config, and workflow modes for Plan/Execute/Verify advances, plan-scoped .plans/ tools, and compaction behavior
• Custom TypeScript extensions — Subagent dispatch, MCP broker tools, web search/fetch/PDF/GitHub access, TODO tracking with sticky widget, compact tool rendering, startup header, status line, and interactive ask_user
• Reusable skills — Agent engineering, TDD, PR review, Playwright browser automation, Jira ticket creation, frontend design, and skill creation, adapted for Pi conventions and GPT-5.x prose
• Delegated subagents — Definitions for focused exploration, fast research, deep research, and review, loaded dynamically by the subagents extension
• Extension dev harness — Colocated TypeScript tests, shared extension helpers, and make typecheck / make test coverage for Pi extension logic

Companion: agent-tools

agent-config configures the agent; agent-tools provides tools for working with AI agents. Two of its tools are explicit integration points for this repo:

• MCP broker — credentials-holding proxy that lets sandboxed agents use external tools without ever holding the secrets themselves. Pairs with the mcp-broker Pi extension and Claude's sandbox-mode gh/git redirection hooks
• Sandbox manager (sb) — provisions and manages a Lima-based Linux VM for isolated agent runs. Pairs with the overrides in claude/sandbox/

Quick Start

Requirements

• Claude Code and/or Pi agent
• Homebrew
• Node.js 24+
• macOS assumed, adaptable for Linux

Setup

git clone git@github.com:averycrespi/agent-config.git
cd agent-config
brew bundle             # install system dependencies on macOS
make install-playwright # for /playwright-cli skill
make stow-claude        # symlink claude/ into ~/.claude/
make stow-pi            # symlink pi/agent/ into ~/.pi/agent/

Development

make install-dev      # install Pi dev dependencies and Husky git hooks
npm run lint          # lint Pi extension TypeScript files
npm run format:check  # check formatting for TS/JS/JSON/Markdown/YAML files
make typecheck        # type-check Pi extension TypeScript files
make test             # run Pi extension unit tests

npm install runs Husky's prepare script, which installs the repo's pre-commit hook. The hook runs lint-staged for staged formatting/lint fixes, then npm run typecheck.

License

• Repository licensed under MIT
• Individual components may have their own licenses