Overview
Another Agent Skills is a framework of 57 composable skills that turn AI coding agents into disciplined senior engineers.
What You'll Find
Getting Started
Installation, setup, and first project walkthrough.
Quick Start Guide
First session walkthrough, how skills activate, common scenarios, and tips.
Lifecycle
The 6-phase development flow: Define, Plan, Build, Verify, Review, Ship.
Skills
Browse all 57 skills with filtering by category and lifecycle phase.
The Harness
6-component harness architecture: instructions, tools, sandboxes, orchestration, guardrails, observability.
Evaluation Guide
How the skill eval system works: format, tiers, adding evals, running tests, trigger accuracy, and coverage checklist.
Philosophy
SOUL.md principles, honest limitations, and what we never do.
Agents
Compatible agents: OpenCode, Claude Code, Cursor, Devin Desktop, and more.
Quick Links
| Current version | v6.1.0 |
| Skills | 57 |
| Guides | 74 |
| License | MIT |
What's New in v6.1.0
- Claude Code Parity — 57 skills auto-install to ~/.claude/skills/ (auto-discovered, same as OpenCode). One command, no per-project setup. Full release notes →
- Enforcement Hooks, Made to Actually Work — commit-approval, pre-flight, and edit-guard now parse Claude Code's real hook JSON and auto-wire into .claude/settings.json via an idempotent merge that never touches your own hooks.
- 19 New Tests — Hook behavior, install idempotency, and EN/ES doc parity all covered — 28/28 test suites passing.
Previous releases
- v6.0.0 — Phase 6: Design Skill Integrity — 17-section DESIGN.md schema, 3-mode design-gate.sh, TDD enforcement with no override, Gate 0 DECISION_APPROVED block. Full release notes →
- v5.0.0 — Phase 4: Docs Honesty Complete — 42 issues fixed across 6 groups, 86 files changed. Zero stale versions, hooks, i18n gaps, or nav inconsistencies. Full release notes →
- v4.2.0 — Phase 3: Output Contracts Complete — All 57 skills now have standardized Output Contracts. Check 16 warnings eliminated (37→0). Full release notes →
- v4.0.0 — Foundation Repair & Critical Stubs — 15 critical stub skills completed. All 57 skills ≤250 lines. Hook simplification (commit-msg v7→v4). Pre-commit v11 (14 gates). Full release notes →
- v3.1.1 — Test Infrastructure — Unified test runner (
tests/run-all.sh), pre-commit Gate 14, name-pairing + new-test enforcement. - v3.1.0 — TDD Enforcement Gate — Commit-msg Gate 4 enforces TDD. Pre-commit v11 with 14 gates. Learn more →
- v3.0.0 — Universal Self-Improvement Loop — Config-driven audit engine (
universal-audit.sh) for any stack. Learn more →