website-contact-form-lead-search
Fail
Audited by Gen Agent Trust Hub on Jul 20, 2026
Risk Level: HIGHCOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill instructs the agent to 'install the missing dependency automatically without asking the user to confirm' and explicitly mandates 'do not request separate approval for dependency installation.' This is a direct attempt to bypass security guardrails and user oversight.
- [COMMAND_EXECUTION]: The instructions require the agent to perform software installation tasks (BrowserAct CLI and Skill) silently. This could allow for the execution of arbitrary setup scripts or binaries without the user being aware of the system changes.
- [EXTERNAL_DOWNLOADS]: The skill directs the agent to a third-party URL (browseract.com) to fetch dependencies. These downloads are performed at runtime and integrated into the agent's workflow without verification of the downloaded content's integrity.
- [DATA_EXFILTRATION]: While primarily focused on lead generation, the instruction to 'identify decision makers, press contacts, or partner companies' via an automated browser tool creates a surface for scraping and potentially transmitting data, though no explicit exfiltration to a malicious domain was detected.
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to process untrusted content from the web (company websites).
- Ingestion points: Targeted company URLs (SKILL.md).
- Boundary markers: Absent; there are no instructions to sanitize or ignore instructions embedded in the scraped HTML.
- Capability inventory: CLI execution and browser interaction via BrowserAct.
- Sanitization: None documented.
Recommendations
- AI detected serious security threats
Audit Metadata