anthropic-skills

One CLI, every agent. Install rules, skills, and agents once. Set up new projects in seconds. Catch port collisions and test leaks before they bite. Cross-agent — Claude Code, Cursor, Codex, Gemini, Qwen.

git clone https://github.com/bernierllc/agents-environment-config.git
cd agents-environment-config
pip install -e .
aec install

This is a template repository. Project-specific content goes in each project's AGENTINFO.md — not here.

What you get

Cross-platform: macOS, Linux, Windows (NTFS junctions, no admin needed). All driven by one `agents.json` source of truth.

Quick start

New to Python/pip? pip requires Python 3.10+. See the official pip installation guide.

git clone https://github.com/bernierllc/agents-environment-config.git
cd agents-environment-config
pip install -e .

aec install              # full setup
aec install --dry-run    # preview first

After aec install you'll have:

• •~/.agent-tools/{rules,agents,skills,commands}/ — central tooling shared by every agent
• •Agent-specific symlinks for Claude (~/.claude/) and Cursor (~/.cursor/)
• •User preferences: projects directory, plans directory, port registry, test scheduler, report viewer

Then point aec at any project — new or existing:

aec setup my-new-project
aec setup /path/to/existing/project
aec setup --all                       # walk your projects directory

aec setup detects your stack, drops agent files, installs lint hooks, registers ports, finds existing skills/agents/rules, and offers to fill in missing git essentials.

Supported agents

AEC handles the format differences between agents automatically. You write rules once; each agent gets them in the format it expects.

All non-Cursor agents read .agent-rules/*.md (plain markdown). Cursor reads .cursor/rules/*.mdc (with YAML frontmatter). AEC keeps both in sync from a single source. See Rules architecture.

Adding another agent is one JSON entry — see Adding agent support.

Solo, team, or org

• •Solo. Clone, aec install, done. Every project picks up a coherent set of rules, agents, skills, and MCP servers.
• •Team. Share .agent-rules/ and .cursor/rules/ across machines. AEC's parity check catches drift between formats.
• •Org. IT/admin teams publish an org config — a single YAML describing required, recommended, blocked, and pinned items, plus default preferences. Users enroll with aec org enroll <path>. AEC never hosts org configs; each org publishes their own. (Phase 1, behind pip install aec[org-configs-preview].) See Org configs for users and Authoring org configs.

Command cheat sheet

# Install / update
aec install                  # full setup
aec install <type> <name>    # one skill / rule / agent
aec update && aec upgrade    # fetch + apply upgrades
aec doctor                   # health check

# Projects
aec setup [path]             # track a project
aec setup --all              # track everything in your projects dir
aec list                     # show installed items
aec search <term>            # search catalog

# Ports
aec ports list | check | register | unregister | validate

# Tests
aec test run [-g]            # local / scheduled cross-project
aec test schedule            # set up automated daily runs
aec test report [-g]         # latest results

# Discovery
aec discover [-g]            # find existing items matching the catalog

Full reference: CLI commands.

Documentation

Using AEC

• •CLI reference
• •Catalog: installing skills, rules, agents
• •Discovery
• •Lint hooks
• •Port registry
• •Test runner & scheduler
• •Git essentials phase
• •`.aec.json` schema
• •Preferences & local config
• •Org configs
• •Raycast integration
• •Rules architecture
• •Troubleshooting

Contributing

• •Architecture
• •Generation pipeline
• •Testing
• •Adding agent support
• •Adding hook language
• •Adding git provider
• •Adding test framework

Repository layout

agents-environment-config/
├── aec/                  # Python CLI package
├── .agent-rules/         # Rules without Cursor frontmatter (generated)
├── .cursor/rules/        # Rules with frontmatter (source of truth)
├── .claude/{agents,skills}/   # Submodules: agency-agents, skills
├── agents.json           # Single source of truth for agent definitions
├── templates/            # Files copied into your projects (CLAUDE.md, etc.)
├── scripts/              # Setup, hooks, validators
├── raycast_scripts/      # Raycast launchers (parity-checked with scripts/)
└── docs/                 # User & contributor docs

Trade-offs

Related repos

• •agency-agents — agent definitions
• •skills — skill definitions

Security

• •Never commit .env files. API keys go in .env (copy from .env.template).
• •Review API key permissions — use least-privilege access.

Hook key repair

Earlier versions of AEC wrote .claude/settings.json with camelCase hook keys (e.g., postToolUse) instead of the PascalCase keys Claude Code requires (PostToolUse). AEC automatically detects and fixes this in all tracked repos when you run aec install, aec update && aec upgrade, or aec doctor. No manual editing needed.