Get started

Overview

TL;DRAnother Agent Skills is a framework of 58 composable skills plus mechanical enforcement (L1 local hooks, L2 a required remote gates check, L3 CODEOWNERS review) that turns AI coding agents into disciplined senior engineers.

What it is

Another Agent Skills is a framework of 58 composable skills that turn AI coding agents into disciplined senior engineers. Most skill libraries sell capability. This one sells discipline you can verify: every task follows a six-phase lifecycle, and the parts that matter are enforced by mechanisms, not by prompts.

Start with /gate. The entry-point skill detects your stack, wires the local hooks, the TDD gate, and the required CI check, and proves the first gate actually fires.

The core idea is simple. A rule that lives only in a file is a suggestion. A rule that lives in a layer is a gate. The framework ships the layers: local git hooks for fast feedback, a required remote gates check for authority, and CODEOWNERS review for the gate configuration itself.

The thesis

  • Capable models still skip tests, review, and context. The gap is process, not intelligence.
  • Skills map to a lifecycle: Define, Plan, Build, Verify, Review, Ship. No phase is optional.
  • Enforcement has three layers. A code change with no matching test is blocked, locally and remotely.
  • The agent never runs git commit or git push. The human does.

At a glance

Fact Value
Skills 58, mapped to the six-phase lifecycle
Guides 151
Harness components 6
Enforcement L1 local hooks, L2 required gates check, L3 CODEOWNERS
License MIT
Install channels git clone, pinned curl bootstrap and npm (all live)
Agents OpenCode first, portable to Claude Code, Cursor, Codex, Gemini CLI, Copilot, and any agent that reads AGENTS.md

Where to go next

Guides

Five short walkthroughs, each with copy-paste commands and the output you should see:

The FAQ answers the common questions directly and cites the facts.

The framework works offline after install, has no external runtime dependencies, and ships no trackers.

Browse the docs

Getting startedInstall Another Agent Skills, run init-agents, and see the local gate block a commit that has no matching test.Your first gated commitA five-minute walkthrough: install Another Agent Skills, run init-agents, watch the local gate block a code change with no matching test, then add the test and commit.Wire the remote enforcementTurn L1 feedback into L2 authority: commit the gates workflow, open a pull request so the gates check reports, preview branch protection with --dry-run, apply it, and verify.Start without git, add it laterUse Another Agent Skills in a folder with no git, then grow the enforcement as layers appear: git init, re-run init-agents for L1, add a GitHub remote, re-run again for L2.Migrate a legacy projectInherited a project that used an older Another Agent Skills layout? Run aas doctor, preview with init-agents --dry-run, then migrate with --repair without losing the team's AGENTS.md.Move to another machineRecreate your setup on a new computer: install the same framework version, run aas install in the project, confirm with aas doctor, and the project works without reconfiguring it.LifecycleThe six-phase development lifecycle: Define, Plan, Build, Verify, Review, Ship. Each phase has an exit criterion and a set of skills.SkillsDetailed reference for the 58 skills: what each one does, when it activates, when to use it and when not to.Enforcement (L1/L2/L3)The three enforcement layers: L1 local hooks, L2 the required remote gates check, and L3 CODEOWNERS review. A code change with no matching test is blocked.AgentsCompatibility matrix and setup per agent: OpenCode, Claude Code, Cursor, Kiro, and any agent that uses git.Distribution and upgradesHow Another Agent Skills reaches users: git clone, the pinned curl bootstrap, the aas CLI, and the npm wrapper, plus how upgrades work.Branch protectionHow the remote enforcement layer works: GitHub branch protection, the required gates check, CODEOWNERS, and the solo and team profiles.FAQDirect answers about installing, upgrading, enforcement, agents, and what the framework does and does not guarantee.

Edit this page