agent-readablelisted
Install: claude install-skill zydo/skills
# agent-readable (Python + TypeScript)
Two sibling libraries, one philosophy: before coding against an unfamiliar API,
**inspect it at runtime** to get the live public surface and any author-supplied
behavioral rules — instead of guessing from stale training data.
| | Python | TypeScript / JS |
|---|---|---|
| Library | `agent-readable` | `agent-readable-ts` |
| Install | `pip install agent-readable` | `npm install agent-readable-ts` |
| One-off, no install | `uvx agent-readable <target>` | `npx agent-readable-ts <target>` |
| Inspect function | `agent_help(target)` | `agentHelp(target)` |
| Author notes method | `__agent_notes__()` classmethod | `agentNotes()` method |
| Custom output method | `__agent_help__()` | `agentHelper()` |
| Protocol / mixin | `AgentReadable`, `AgentReadableMixin` | `AgentHelper`, `AgentNoter` interfaces |
***
## When to activate
Call the inspect function before writing code against an API where guessing could
plausibly be wrong — concretely, when:
* The target is **project-local or third-party code you don't already know cold**.
* It's **stateful, version-sensitive, poorly documented, has lifecycle or ordering
constraints, or exposes several similar entrypoints** that are easy to confuse.
* You're **adding or changing a public API** meant for reuse by humans or agents.
* The user mentions `agent_help`, `agentHelp`, `__agent_notes__`, `agentNotes`,
or the `agent-readable` / `agent-readable-ts` library.
## When not to activate
Skip it for