The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill directs the agent to install the @jaw.id/cli package from the NPM registry and references the Foundry cast tool for calldata encoding.
[COMMAND_EXECUTION]: The skill instructions involve executing the jaw CLI binary for wallet management, configuration, and RPC calls. It also describes the setup of an MCP server (jaw mcp) which runs as a persistent subprocess within the AI agent environment.
[DATA_EXFILTRATION]: The skill performs network requests to api.justaname.id for ENS resolution and communicates with keys.jaw.id for session management and transaction signing. These operations are functional requirements for the author's service.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection in rules/ens-resolution.md where the agent fetches external data used to define transaction parameters.
Ingestion points: Fetches ENS records from https://api.justaname.id/ens/v1/subname/records (rules/ens-resolution.md).
Boundary markers: None explicitly defined for the API response, though specific parsing rules are provided.
Capability inventory: Sensitive wallet capabilities including wallet_sendCalls and eth_sendTransaction (rules/api-reference.md, rules/transactions.md).
Sanitization: Instructions provide specific logic for extracting addresses from the coins array based on chain IDs and require showing the resolved address to the user, which mitigates the risk of processing arbitrary data.

jaw-cli