landing-page-intel-extractor
Pass
Audited by Gen Agent Trust Hub on Jul 10, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection due to its core functionality of fetching and parsing arbitrary external HTML.
- Ingestion points: The skill retrieves content from user-provided URLs and automatically discovered paths (e.g.,
/customers,/pricing,/vs/) using theWebFetchtool. - Boundary markers: The instructions do not define clear delimiters for the fetched content, nor do they include specific directives for the agent to ignore or isolate potential instructions embedded within the retrieved HTML.
- Capability inventory: The skill has access to read local configuration files (
clients/<client>/context.md), write new files to the filesystem (clients/<client-name>/intel/), and perform additional network operations (WebSearch). - Sanitization: There is no evidence of sanitization, filtering, or validation of the fetched HTML content before it is processed by the LLM for signal extraction.
Audit Metadata