The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The skill requires users to maintain a local file named coinpilot.json that stores high-sensitivity secrets in plaintext, including the Coinpilot API key, a Privy user ID, and up to ten wallet private keys.
[DATA_EXFILTRATION]: The Node.js execution script (scripts/coinpilot_cli.mjs) transmits unencrypted wallet private keys and API credentials to the remote domain api.coinpilot.bot via HTTP headers (x-wallet-private-key) and request body parameters (primaryWalletPrivateKey, followerWalletPrivateKey). This behavior exposes the user's full wallet control to the vendor's infrastructure and any potential network intermediaries.
[COMMAND_EXECUTION]: The skill uses the system's node binary to run a local JavaScript file (scripts/coinpilot_cli.mjs) which contains the core logic for authentication, API interaction, and trading execution.
[PROMPT_INJECTION]: The skill possesses an attack surface for indirect prompt injection by processing untrusted data from external lead wallet metrics and category rankings.
Ingestion points: Data returned from the Coinpilot API via endpoints like /lead-wallets/metrics/wallets/:wallet and /lead-wallets/metrics/categories/:category is ingested and presented to the agent.
Boundary markers: The skill instructions do not define specific delimiters or security boundaries to isolate external API data from the agent's internal logic.
Capability inventory: The agent has the capability to execute shell commands (via node), read local configuration files, and perform financial transactions (trading) based on processed data.
Sanitization: There is no evidence of sanitization or structural validation to prevent malicious instructions embedded in lead wallet metadata or metrics from influencing the agent's behavior.

coinpilot-hyperliquid-copy-trade