Skill for AI assistants

The Immunity SDK ships with a packaged agent skill, a single SKILL.md file an AI coding tool can read to understand the API surface, the empirical gotchas, and the right way to wire check() into your agent. It cuts the round-trips needed to get a correct integration on the first try.

The skill is part of immunity-protocol/skills, a public repository of empirical skills for any tool implementing the agentskills.io spec.

npx skills add immunity-protocol/skills --skill immunity

The CLI installs to ~/.agents/skills/immunity/ and symlinks into your agent's skill directory automatically. Works for Claude Code, Cursor, Codex, Cline, Gemini CLI, and any tool that scans the standard skill paths.

Or download manually

Pick whichever tool you use:

Claude Code

mkdir -p ~/.claude/skills/immunity
curl -fsSL https://raw.githubusercontent.com/immunity-protocol/skills/main/skills/immunity/SKILL.md \
  > ~/.claude/skills/immunity/SKILL.md

Restart Claude Code and the skill activates whenever it detects an Immunity context in your project.

Cursor

mkdir -p .cursor/rules
curl -fsSL https://raw.githubusercontent.com/immunity-protocol/skills/main/skills/immunity/SKILL.md \
  > .cursor/rules/immunity.mdc

Cursor reads .cursor/rules/*.mdc per workspace. Add the file to your project root.

Codex

Paste the SKILL.md verbatim into the system-prompt include of your Codex setup. Most teams keep these under prompts/skills/immunity.md and load them at boot.

Cline / Gemini CLI / others

Drop the file at the standard agentskills path for your tool. Most tools follow the same layout: a per-skill directory with the SKILL.md inside, scanned at startup or on workspace change.

What's in the skill

  • The full SDK API surface, every public method, every option, every error class.
  • The three-tier lookup walkthrough with concrete latency and cost numbers.
  • Configuration table mirrored from the SDK README so the skill stays in sync.
  • The five antibody types with real catches drawn from the demo's incident catalog.
  • Empirical gotchas: peer-dep flag, AXL-as-separate-daemon, port 5678, TEE key rotation, etc.
  • A distilled key-rules checklist your assistant can use as a self-check.

Why bother

Without the skill, your assistant guesses based on outdated training data, misnames methods, hallucinates options, and writes setup boilerplate that does not match the SDK's actual peer-dep requirements. With the skill in place, code-completion and code-generation produce idiomatic Immunity calls on the first try.

Source

The skill lives at github.com/immunity-protocol/skills/blob/main/skills/immunity/SKILL.md. It is updated alongside SDK releases. Pin to a tag in your install if you need stability across SDK upgrades.