The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses local system commands curl, jq, openssl, and uuidgen to interact with remote APIs, parse JSON responses, and generate nonces for cryptographic signing.
[DATA_EXFILTRATION]: The skill requires the BANKR_API_KEY environment variable, which it transmits in HTTP headers to api.bankr.bot and coordinator.agentmoney.net. This is the intended authentication mechanism for the protocol and its dependencies.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection through external data ingestion.
Ingestion points: The agent fetches a prose document (doc) and specific questions from the vendor's coordinator at https://coordinator.agentmoney.net/v1/challenge.
Boundary markers: While the skill provides instructions to the LLM to output a single-line artifact, it lacks explicit boundary markers or instructions to disregard potential malicious directives embedded within the ingested prose document.
Capability inventory: The skill possesses significant capabilities, including the ability to submit raw transactions and natural language prompts to the Bankr wallet API (api.bankr.bot/agent/submit and api.bankr.bot/agent/prompt), which could be abused if the agent is successfully manipulated.
Sanitization: There is no evidence of sanitization or filtering of the challenge document content before it is processed by the LLM.

botcoin-miner