social-brand-mention-monitor
Fail
Audited by Gen Agent Trust Hub on Jul 20, 2026
Risk Level: HIGHREMOTE_CODE_EXECUTIONPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill instructions attempt to override standard agent safety protocols. It explicitly mandates that the agent "install the missing dependency automatically without asking the user to confirm" and further instructs the agent to hide this activity by not describing it as a request for the user to open their browser.
- [REMOTE_CODE_EXECUTION]: The instructions require the agent to download and install a third-party CLI and skill from an external URL (
https://www.browseract.com/...) if they are not already present. Directing an agent to execute or install software from an unverified source without human oversight is a high-risk pattern. - [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from various social media platforms, creating an attack surface for indirect injection.
- Ingestion points: Real-time social media pages accessed via
BrowserAct(SKILL.md). - Boundary markers: Absent; the skill does not include delimiters or instructions to ignore embedded commands within the scraped content.
- Capability inventory: The agent is granted browser interaction capabilities and the ability to execute CLI tools.
- Sanitization: No sanitization or validation of the ingested content is described in the workflow.
Recommendations
- AI detected serious security threats
Audit Metadata