ℹ
Purpose & Capability
Name/description (registering Handshake-like domains on Ethereum via Impervious Domains contracts) aligns with requiring an Ethereum RPC endpoint, but the skill claims it will mint ERC‑721 NFTs yet does not declare how transactions will be signed or how the user's wallet is supplied.
!
Instruction Scope
SKILL.md instructs the agent to follow a full procedure and points to `{baseDir}/../../skills.md` for ABIs, addresses, and safety constraints — those files are not included. The instructions implicitly require sending signed transactions (commit/reveal) and waiting for confirmations but give no guidance on where signing keys or user wallet access come from. Referencing files outside the skill bundle is scope creep and could cause the agent to read arbitrary host files.
✓
Install Mechanism
No install spec and no code files — lowest-risk delivery method. However, being instruction-only means the runtime behavior depends entirely on what the agent is told to do (e.g., network calls to the RPC).
!
Credentials
Only ETHEREUM_RPC_URL is declared, but minting on-chain requires transaction signing (private key, hardware wallet, or user-signed payloads). The absence of any declared signing credential or explicit user-interactive signing workflow is a mismatch and could lead to the skill asking for private keys or other secrets at runtime. Also ETHEREUM_RPC_URL may embed API keys — the skill provides no guidance on acceptable RPC providers or scopes.
✓
Persistence & Privilege
always is false, no config paths requested, and no install steps that modify system or other skills. Persistence/privilege requirements appear minimal.
What to consider before installing
This skill is ambiguous rather than clearly malicious, but proceed cautiously. Before installing or using it, ask the author to: (1) provide the full ABIs and contract addresses used (not a relative `{baseDir}` reference), (2) explicitly describe how transactions are signed (do not share private keys — prefer user-side signing or hardware wallet), and (3) show the exact RPC endpoints and explain any API keys. Never paste your private key or seed phrase into the agent; instead, require the agent to prepare unsigned transactions for you to sign offline or via your wallet. If you must test, use a fresh ephemeral wallet with minimal funds on a testnet first and verify every transaction payload and destination contract address. If the skill asks to read local files outside its bundle or to accept a private key, decline until the above are clarified. If the skill's source or homepage is unknown, prefer not to grant sensitive access (private keys, full-access RPC endpoints) until the code or a trustworthy origin is available.
