agent-md
Production-grade agent directives for autonomous coding agents: Claude Code, Codex, Cursor, Windsurf, Aider.
agent-md is a portable agent skill that installs production-grade directives and enforcement contracts for autonomous coding agents including Claude Code, Codex, Cursor, Windsurf, and Aider. It deploys a single source-of-truth rules file (AGENT.md), repository-local hooks, persistent task state in memory files, and helper scripts so agents follow consistent verification workflows. The skill moves reliability out of advisory prose and into enforceable artifacts like type-checks, tests, visual evidence, and state updates. Developers install it once per project to ensure agents document progress, verify UI changes with screenshots, and block destructive commands.
Key Features
Use Cases
- 01Enforcing type-checks and test runs before an autonomous agent completes a coding task
- 02Maintaining a durable progress log across agent sessions so work history persists in memory/progress.md
- 03Requiring visual evidence (screenshots and annotations) for UI feature work to prevent unverified claims
- 04Blocking destructive bash commands like rm -rf before they execute via pre-tool-use hooks
- 05Standardizing agent behavior across multiple autonomous coding tools in the same repository
- 06Ensuring agents document stack, tooling, and known gotchas in memory/ files for human handoff
Related Skills
View moresuperpowers
An agentic skills framework & software development methodology that works.
skills
Skills for Real Engineers. Straight from my .agents directory.
skills
Public repository for Agent Skills
ponytail
Makes your AI agent think like the laziest senior dev in the room. The best code is the code you never wrote.
agent-md — FAQ
What is agent-md?+
agent-md is an agent skill that installs portable rules files, enforcement hooks, and state management for autonomous coding agents. It shifts reliability from advisory instructions to checked artifacts like tests, type-checks, and visual evidence.
How do I install agent-md in my project?+
Run curl -sL https://raw.githubusercontent.com/iamfakeguru/agent-md/main/install.sh | bash from your project directory. The installer scaffolds AGENT.md, hooks, memory files, and helpers for Claude Code, Codex, Cursor, and Windsurf by default.
Which AI coding agents does agent-md support?+
agent-md supports Claude Code, Codex, Cursor, Windsurf, and Aider. Claude Code and Codex receive hard-block enforcement hooks; Cursor and Windsurf get project rules plus optional git-hook fallback.
Does agent-md require API keys or configuration?+
No API keys are required. Optional configuration is done via agent-md.toml to declare typecheck, lint, test commands, and visual evidence requirements for deterministic verification.
Is agent-md free to use?+
Yes, agent-md is open source and released under the MIT license.
Can agent-md prevent all destructive commands?+
Bash safety hooks provide guardrails for Claude Code and Codex but are not a full sandbox. Git hooks can be bypassed with --no-verify, and enforcement depends on the host agent exposing hook events.
How do I install agent-md?+
Open the source repository on GitHub and follow its README. agent-md is a skill — MCP Agents Market links you directly to the official repo.
Is agent-md free?+
agent-md is an open-source project hosted on GitHub. Check the repository for its license and any usage requirements.