local-business-prospecting
Pass
Audited by Gen Agent Trust Hub on Nov 10, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill uses a local bash script
scripts/crawlora.shto interface with the Crawlora API. The script is designed with significant security controls, including a strict whitelist for API paths and HTTP methods, validation of the API key format, and the use of a temporary curl configuration file to prevent credential leakage in process listings. It also sanitizes input arguments to prevent local file disclosure through curl's file-read syntax. - [INDIRECT_PROMPT_INJECTION]: The prospecting workflow ingests data from untrusted third-party sources, including Google Maps, Yelp, and business websites. 1. Ingestion points: Data retrieved via the
/web/scrapeand/extractendpoints from the Crawlora API is incorporated into the agent's context. 2. Boundary markers: The skill does not define explicit delimiters for separating tool output from the system prompt. 3. Capability inventory: The agent can perform subsequent network requests to the Crawlora API and generate local files (e.g., CSV). 4. Sanitization: Instructions specifically direct the agent to only extract public business contact details and avoid inferring information, serving as a procedural guardrail against processing malicious content.
Audit Metadata