The Agent Skills Directory

[SAFE]: The skill implements a defensive helper script scripts/crawlora.sh for API interactions.
The script validates the format of the CRAWLORA_API_KEY to prevent injection into configuration files.
It uses a temporary configuration file with restricted permissions (chmod 600) to pass the API key to curl, preventing exposure in process listings.
It strictly whitelists allowed API routes and restricts operations to GET requests to minimize the attack surface.
It sanitizes query parameters by rejecting the @ character to prevent local file inclusion vulnerabilities via curl's data flags.
[INDIRECT_PROMPT_INJECTION]: The skill processes data from external sources (Fortune and Forbes rankings) which constitutes an indirect injection surface.
Ingestion points: Data is retrieved from api.crawlora.net via scripts/crawlora.sh and processed by the agent.
Boundary markers: The instructions advise preserving source metadata but do not implement explicit delimiters for the external content.
Capability inventory: Capabilities are limited to authenticated GET requests to a hardcoded vendor API; the skill lacks file system write or arbitrary code execution permissions.
Sanitization: The helper script performs path whitelisting and URL-encodes query parameters to ensure integrity.

company-ranking-research