brand-mention-research
Pass
Audited by Gen Agent Trust Hub on Oct 20, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from platforms like Reddit, TikTok, and YouTube, which could contain adversarial instructions.
- Ingestion points: Social media content is retrieved through the Crawlora API via
scripts/crawlora.sh. - Boundary markers: The instructions do not define specific delimiters to separate external content from the agent's system prompt.
- Capability inventory: The skill's capabilities are restricted to network GET requests; no file-writing or code execution is performed on the results.
- Sanitization: API parameters are URL-encoded by the helper script to prevent injection during the request phase.
- [DYNAMIC_EXECUTION]: The
scripts/crawlora.shscript dynamically buildscurlcommands based on agent input. - Evidence: The script uses shell variables to construct URLs and query strings for the Crawlora API.
- Mitigations: The script enforces a hardcoded vendor base URL, validates the request path against a strict regex-based allowlist, and URL-encodes all query parameters.
Audit Metadata