playbook-pricing-page
Pass
Audited by Gen Agent Trust Hub on Aug 23, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface because it ingests untrusted content from external websites to generate structured records and text for email campaigns.
- Ingestion points: The
Visible page textparameter in the extractor prompt (SKILL.md) receives raw data scraped from arbitrary domains. - Boundary markers: The prompt uses JSON response formatting and specific rules (e.g., "Never invent a number") to constrain output, but lacks robust delimiters between instruction and data.
- Capability inventory: The skill generates a
pricing_linespecifically intended for use in outbound communications, which could be manipulated by content on the target page. - Sanitization: The documentation (clay-table.md and clay-workflow.md) specifies a 'Price Assertion' check that validates whether extracted prices and plan names appear verbatim in the source text, which serves as a mitigation against fabrication.
Audit Metadata