The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill fetches data from an external source (Facebook profiles via the Frevana API) which could contain adversarial content or instructions intended to override the agent's behavior.
Ingestion points: The script scripts/get_facebook_profile.sh fetches JSON data from https://ai-factory.frevana.com/service/facebook-profile and passes it back to the agent.
Boundary markers: There are no explicit delimiters or warnings to the agent to ignore instructions embedded within the fetched profile data.
Capability inventory: The skill includes a bash script that uses curl for network requests and python3 for data processing. It also supports writing the output to a user-specified file path via the --output flag.
Sanitization: The script includes a Python block that validates the response is valid JSON before returning it to the agent.
[COMMAND_EXECUTION]: The skill relies on a local bash script (scripts/get_facebook_profile.sh) which invokes curl and python3. The scripts are used safely, with arguments passed to Python via environment variables to prevent shell injection.

facebook-profile