The Agent Skills Directory

[DATA_EXPOSURE_AND_EXFILTRATION]: The skill is instructed to read an API key from a specific hidden file in the user's home directory (~/.deepapi/env). While intended for authentication to the described service, accessing files outside the project directory is a sensitive operation.
[DATA_EXPOSURE_AND_EXFILTRATION]: The skill performs network requests to an external domain (deepapi.co) using curl, transmitting the retrieved API key in the Authorization header.
[INDIRECT_PROMPT_INJECTION]: The skill processes research results from an external API, which could contain instructions that influence the agent's behavior during the verification process.
Ingestion points: External research data returned from the POST /v1/research/deep endpoint (SKILL.md).
Boundary markers: None identified; the skill does not specify delimiters or warnings to ignore instructions embedded in the API responses.
Capability inventory: The skill uses curl for network operations and writes verification results to the local file system (e.g., docs/evals/) (SKILL.md).
Sanitization: No sanitization or validation mechanisms are described for the data received from the external research service.
[COMMAND_EXECUTION]: The skill provides a bash snippet using curl and uuidgen to interact with the DeepAPI service, which the agent is expected to execute.

risky-changes