The Agent Skills Directory

[COMMAND_EXECUTION]: The skill employs a secure helper script, scripts/crawlora.sh, to manage network requests to the vendor's API. The script implements several defensive measures:
Credential Security: It uses a temporary, restricted-access curl configuration file to pass the x-api-key header, ensuring the API key is never exposed in process lists or shell history.
Path Validation: A strict allow-list and regular expression matching are used to restrict API calls to specific, intended Facebook Marketplace and Page endpoints.
Injection Prevention: The script rejects potentially dangerous characters like @ in query parameters and streams POST data via stdin, effectively preventing curl from being used to read or exfiltrate local system files.
[INDIRECT_PROMPT_INJECTION]: The skill processes external content from Facebook listings and public pages, which presents a standard surface for indirect prompt injection.
Ingestion points: External data enters the system through the JSON responses fetched by scripts/crawlora.sh from the Crawlora API.
Boundary markers: The data is provided in a structured JSON format, providing a clear boundary between instructions and data.
Capability inventory: The skill is limited to making requests to the api.crawlora.net domain.
Sanitization: The skill retrieves and returns raw public data as-is, relying on the agent's internal guardrails for processing untrusted content.

facebook-research