The Agent Skills Directory

[DATA_EXFILTRATION]: The skill attempts to access a sensitive file path at ~/.deepapi/env to retrieve environment variables and API keys. Access to hidden configuration files in the user's home directory poses a data exposure risk.
[DATA_EXFILTRATION]: Sensitive data obtained from the local environment file is transmitted to an external domain (deepapi.co) as an authentication token in the request headers.
[EXTERNAL_DOWNLOADS]: The skill makes several network requests to the domain deepapi.co to perform web searches and scrape external websites. This domain is not identified as a well-known or trusted service.
[COMMAND_EXECUTION]: The skill instructs the agent to execute various shell commands, including source for loading local configuration, curl for interacting with the external API, and uuidgen for creating idempotency keys.
[PROMPT_INJECTION]: The skill possesses a vulnerability surface for indirect prompt injection due to its automated scraping capabilities.
Ingestion points: External data enters the agent's context through web scraping (/v1/scrape/website) and social media searches (/v1/scrape/twitter/search) as defined in SKILL.md.
Boundary markers: The skill does not define any boundary markers or instructions to ignore potential commands embedded within the scraped content.
Capability inventory: The agent can perform network requests via curl and generate markdown responses based on the external data it retrieves.
Sanitization: There is no mention of sanitizing, filtering, or validating the integrity of the content retrieved from external websites before it is processed by the agent.

online-shopping