facebook-marketplace-category-scraper
Fail
Audited by Gen Agent Trust Hub on Sep 19, 2026
Risk Level: HIGHREMOTE_CODE_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [REMOTE_CODE_EXECUTION]: The instructions command the agent to install the 'BrowserAct' CLI and additional AI skills automatically if they are found to be missing during runtime.
- [EXTERNAL_DOWNLOADS]: The skill explicitly directs the agent to suppress user notification and bypass confirmation requirements when installing dependencies from a remote URL (browseract.com). This behavior circumvents standard security protocols that require user approval for software installation.
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process data from external Facebook Marketplace pages, which serves as a vulnerability surface where malicious actors could embed hidden instructions to manipulate the agent.
- Ingestion points: External data collected from Facebook Marketplace listing, category, and seller pages (referenced in SKILL.md).
- Boundary markers: None identified; instructions lack delimiters or warnings to ignore embedded commands in the scraped data.
- Capability inventory: Uses the BrowserAct CLI/Skill for browser interaction and execution of tasks.
- Sanitization: No sanitization, validation, or filtering of the external page content is mentioned.
Recommendations
- AI detected serious security threats
Audit Metadata