The Agent Skills Directory

[DATA_EXFILTRATION]: The skill searches for personal contact information within the ~/contacts.json file. This data access is restricted to the local environment and is consistent with the skill's core functionality, with no network transmission of the data detected.
[PROMPT_INJECTION]: The skill processes data from a local JSON file, which constitutes an indirect prompt injection attack surface.
Ingestion points: Data enters the agent's context from ~/contacts.json through the jq search command.
Boundary markers: The instructions do not specify any delimiters or warnings to ignore commands embedded in the contact data.
Capability inventory: The skill provides data to tools such as zoom-meeting, whatsapp, and gmail, which possess network and messaging capabilities.
Sanitization: The skill does not implement sanitization or filtering of the content retrieved from the contact file.
Mitigation: The risk is addressed by a mandatory confirmation step, requiring the agent to verify the retrieved information with the user before proceeding with any action.

get-contact