Intuition Agent Skills

Agent skills for Claude Code, Codex, and compatible AI agents. Teaches agents to correctly interact with the Intuition Protocol on-chain.

Skills

Skill	Description	Install
intuition	Canonical reference for producing correct Intuition Protocol transactions -- ABIs, encoding, addresses, value calculations	npx skills add 0xintuition/agent-skills --skill intuition
erc8004-agent-layer	Canonical ERC-8004 identity, classification, trust-provider, and mutable-assessment planning with deterministic registry guardrails	npx skills add 0xintuition/agent-skills --skill erc8004-agent-layer

Quick Start

# Install all skills
npx skills add 0xintuition/agent-skills

# Install a specific skill
npx skills add 0xintuition/agent-skills --skill intuition
npx skills add 0xintuition/agent-skills --skill erc8004-agent-layer

Once installed, skills are available in your agent's session. Use /intuition for generic protocol operations. Use /erc8004-agent-layer to produce the ERC-8004 semantic plan, then start protocol preparation only in a new user turn after reviewing that plan.

What These Skills Do

Intuition runs on an L3 chain that isn't indexed by Etherscan. LLMs can't discover the ABIs, and they make consistent mistakes with the V2 contract interface (bytes32 IDs, batch-only creation, bonding curves). These skills fill those blind spots with verified, canonical knowledge.

The core intuition skill can produce unsigned transaction parameters after the relevant preparation gates. The erc8004-agent-layer skill produces only a semantic plan and stops before transaction preparation. Wallet infrastructure and signing remain the builder's responsibility.

Structure

agent-skills/
├── skills/
│   ├── intuition/               # Core Intuition Protocol skill
│   └── erc8004-agent-layer/     # ERC-8004 partner integration skill
├── .claude-plugin/
│   └── marketplace.json  # skills.sh marketplace manifest
├── CLAUDE.md             # Repo-level agent instructions
├── README.md             # This file
└── LICENSE

Adding Skills

See CONTRIBUTING.md for guidelines on adding new skills to this repo.

Testing

Testing is documented in TESTING.md, including:

• Layer A deterministic calldata checks (scripts/pass2-calldata-verification.sh)
• Layer A.5 edge-case RPC checks (scripts/pass2-edge-case-tests.sh)
• Layer B prompt suites for autonomous consumption and on-chain integration (tests/prompts/)

Releases

Published users should treat main as a moving branch, not a stable channel.

• Pin production installs to a Git tag or commit SHA.
• metadata.version in skills/intuition/SKILL.md is updated only in a release PR, not in ordinary fix branches.
• Release policy, semantic versioning rules, and the publish checklist live in RELEASING.md.
• User-visible release notes live in CHANGELOG.md.

References

• Intuition Protocol
• Agent Skills Specification
• skills.sh -- skill discovery and leaderboard
• Intuition V2 Contracts

License

MIT