The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill communicates with the official Crawlora API (api.crawlora.net) to fetch product details, reviews, and search-box suggestions. This is a vendor-owned resource corresponding to the skill author and is used as intended for data retrieval.
[COMMAND_EXECUTION]: The skill uses a bash helper script (scripts/crawlora.sh) to invoke curl for API requests. The script is highly hardened and includes multiple security best practices:
Validates the CRAWLORA_API_KEY environment variable format using a strict character allowlist ([A-Za-z0-9._-]) to prevent shell injection or configuration smuggling.
Protects the API key from exposure in process listings by writing it to a temporary configuration file created with mktemp and restricted permissions (chmod 600).
Implements a rigid allowlist for API paths using case patterns and anchored regular expressions to ensure the helper only accesses documented routes.
Uses curl -G combined with --data-urlencode to securely handle query parameters, preventing argument injection or malformed URL construction.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from an external source (Macy's product catalog and reviews).
Ingestion points: Data is ingested from the api.crawlora.net endpoint in JSON format.
Boundary markers: None present; the raw API output is passed to the agent.
Capability inventory: The skill is limited to performing network GET requests via the verified helper script.
Sanitization: No explicit sanitization is performed on the product descriptions or reviews before they are returned to the agent. While the risk of an agent obeying instructions hidden in a product review exists, the structured nature of the data and the lack of high-privilege write capabilities in this skill make the risk negligible.

macys-research