book-market-positioning
Pass
Audited by Gen Agent Trust Hub on Oct 20, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill uses a helper script (
scripts/crawlora.sh) to perform API requests. This script implements several security best practices, including: validating the API key format to prevent header injection; restricting operations to the GET method; enforcing a strict whitelist of allowed API paths via regular expressions; and blocking the@character in query parameters to prevent curl's local file disclosure feature. - [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external content such as book descriptions and reader reviews that could contain malicious instructions.
- Ingestion points: Metadata and review text fetched from Goodreads, Apple Books, and Audible via the Crawlora API.
- Boundary markers: Explicit instructions are provided to the agent to treat fetched text as untrusted evidence and to ignore any instructions found within that content.
- Capability inventory: Execution of a local shell script restricted to specific vendor API endpoints.
- Sanitization: The helper script URL-encodes all parameters and validates paths against a permitted catalog to prevent misuse.
Audit Metadata