The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses a local bash script (scripts/crawlora.sh) to perform API requests. The script implements robust security controls, including path validation to reject suspicious characters (e.g., .., //, ?, #) and a whitelist that restricts execution to specific API routes.
[EXTERNAL_DOWNLOADS]: The skill retrieves fashion data from the official Crawlora API (api.crawlora.net). This communication is restricted to the vendor's own infrastructure and is consistent with the skill's primary purpose.
[DATA_EXFILTRATION]: Authenticates to the vendor's API using a CRAWLORA_API_KEY provided via the environment. The helper script manages this key securely by writing it to a temporary configuration file with restricted permissions (600) and utilizing curl's configuration mechanism to keep the credential out of the process list.
[INDIRECT_PROMPT_INJECTION]: The skill processes data from an external source.
Ingestion points: Data enters the agent context as JSON output from scripts/crawlora.sh (fetched from api.crawlora.net).
Boundary markers: None explicitly present in the skill instructions to delimit or label the external API content.
Capability inventory: Subprocess execution via scripts/crawlora.sh and general file/command capabilities defined by the agent's environment.
Sanitization: Outgoing parameters are sanitized via URL-encoding in scripts/crawlora.sh, but no specific validation or filtering is performed on the incoming API response content before processing.

zalando-research