analytics-privacy
Pass
Audited by Gen Agent Trust Hub on Oct 2, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill analyzes external website behavior by ingesting untrusted data through a headless browser. A malicious site could attempt to influence the agent by including specific instructions in its content or metadata.
- Ingestion points: Data is ingested from external URLs via page.goto and network interception in scripts/observe-runtime.mjs.
- Boundary markers: Findings are reported as structured notes, but the skill does not use specific delimiters to segregate external data within the agent prompt.
- Capability inventory: The skill can drive a browser, take screenshots, and intercept network traffic. It lacks capabilities for unauthorized file system modifications or host-level command execution.
- Sanitization: The captureState function in scripts/observe-runtime.mjs attempts to redact values that resemble security tokens or credentials based on string patterns.
Audit Metadata