The Agent Skills Directory

[DATA_EXFILTRATION]: The skill transmits trade metadata and cryptographic signatures to the vendor's API at bankrsignals.com. This is the core functionality required to build a verified trading track record and does not involve exfiltrating sensitive local files or unauthorized credentials.
[COMMAND_EXECUTION]: The provided publish-signal.sh script executes node to generate EIP-191 signatures using the agent's private key. The script dynamically constructs the signing command from trade parameters, which is a standard implementation for this type of integration utility.
[EXTERNAL_DOWNLOADS]: The skill refers the agent to bankrsignals.com/heartbeat.md for periodic checklist updates. This is a documented interaction with the vendor's established infrastructure to ensure the agent maintains synchronized state with the platform.
[PROMPT_INJECTION]: The skill ingests external signal data that includes natural language fields such as reasoning, creating a surface for indirect prompt injection.
Ingestion points: Signal data is retrieved from the /api/feed and /api/signals endpoints as part of the heartbeat routine.
Boundary markers: None explicitly defined; the agent processes descriptions directly from the JSON API response.
Capability inventory: The skill uses curl, bash, node, and python3 to interact with the service and local environment.
Sanitization: The heartbeat example uses python3 to safely parse the structured JSON response, but agents consuming the natural language reasoning field should handle that content as untrusted input.

bankr-signals