The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes a helper script scripts/crawlora.sh which executes the curl command to interact with the Crawlora API. The script implements significant security hardening: it validates the format of the CRAWLORA_API_KEY to prevent shell injection, uses a temporary configuration file with restricted permissions to pass sensitive headers (preventing them from appearing in the process list), and uses strict regex-based path validation to ensure only authorized Zappos API endpoints can be called.
[EXTERNAL_DOWNLOADS]: The skill makes network requests to api.crawlora.net to retrieve product information. This domain is the official infrastructure for the author, crawlora-org, and is used as the primary data source for the skill's functionality.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from Zappos, such as product descriptions and customer reviews, which could theoretically contain malicious instructions.
Ingestion points: Product search results, brand catalog data, and detailed product responses (including reviews and fit feedback) described in SKILL.md and reference/endpoints.md.
Boundary markers: The data is handled as structured JSON, providing basic structural separation, though no specific instructions are given to the agent to disregard embedded commands in the text fields.
Capability inventory: The skill's capabilities are confined to performing scoped network requests via curl to the vendor's API and does not have the ability to write files or execute arbitrary code.
Sanitization: The helper script explicitly blocks the @ character in query arguments to prevent curl from reading local files and handles POST bodies via stdin to avoid file disclosure vulnerabilities.

zappos-research