Agent Rules and Context Guidelines
The toolbelt provides standardized prompt rules, context manifests, and executable skills for multi-agent engineering workflows.
Rule Manifests
| Manifest | Target Runtime | Primary Scope | Guide |
|---|---|---|---|
AGENTS.md | Universal (All Agents & IDEs) | Git Flow, APIs First, SOLID modularity, polyglot matrix, and anti-test-theatre verification. | Universal Agent Rules |
CLAUDE.md | Claude Code | Runtime execution tenets, skill bindings, and deterministic verification links. | Claude Code Rules |
GEMINI.md | Google Antigravity & Gemini CLI | Tool orchestration, workspace discipline, and simulation harness requirements. | Antigravity & Gemini Rules |
.agents/skills/ | Antigravity, Claude Code, Cursor | Executable skills for project scaffolding, archetype enforcement, harness generation, and 6-focus code reviews. | Toolbelt Skills Manifests |
Runtime Integration Matrix
Repository Root Integration
Link AGENTS.md into the root of any software repository that embeds the toolbelt as a submodule:
bash
ln -sf .toolbelt/rules/AGENTS.md AGENTS.mdLocal Skill Registry Installation
Install the executable skills into your local agent directories (~/.gemini/skills/, ~/.claude/skills/) by running:
bash
bash .toolbelt/scripts/install.sh