The Agent Skills Directory

[SAFE]: The helper script scripts/crawlora.sh implements robust defensive measures:
Credential Security: The script validates the format of the CRAWLORA_API_KEY to prevent character injection and stores it in a temporary curl configuration file with restricted permissions (chmod 600). This prevents the key from appearing in process listings.
Path and Route Security: The script enforces a strict allow-list for API routes using case statements and anchored regular expressions, ensuring the agent cannot access unauthorized endpoints.
Injection and Disclosure Prevention: The script explicitly rejects query arguments containing the @ character to prevent curl from reading local files. For POST requests, it streams the body via stdin to avoid similar file-disclosure risks.
[INDIRECT_PROMPT_INJECTION]: The skill processes JSON data from an external API, creating a potential surface for indirect prompt injection if the API content were compromised.
Ingestion points: Data returned from the Crawlora API (api.crawlora.net) via the scripts/crawlora.sh script.
Boundary markers: None present in the instructions.
Capability inventory: The skill is restricted to making network requests to a specific set of Lululemon research endpoints.
Sanitization: The helper script sanitizes the request path and parameters, but the data returned by the API is passed to the agent as raw JSON without content-level sanitization.

lululemon-research