observability

Pass

Audited by Gen Agent Trust Hub on Mar 10, 2026

Risk Level: SAFEPROMPT_INJECTIONDATA_EXFILTRATIONCOMMAND_EXECUTION
Full Analysis
  • [PROMPT_INJECTION]: The skill possesses an indirect prompt injection attack surface because it ingests data from external sources.\n
  • Ingestion points: Consumes browser console logs, network request details, and backend application logs via browser-devtools-cli and node-devtools-cli in SKILL.md.\n
  • Boundary markers: The instructions do not specify the use of delimiters or 'ignore' instructions when handling this external data.\n
  • Capability inventory: The skill can navigate the browser, interact with UI elements, and modify backend execution via tracepoints.\n
  • Sanitization: There is no explicit logic described for sanitizing or filtering logs or request data before they enter the agent's context.\n- [DATA_EXFILTRATION]: The skill configures telemetry exporters to send data to external platforms like Honeycomb and Jaeger. It uses standard placeholders for API keys (e.g., YOUR_API_KEY), avoiding the exposure of actual credentials.\n- [COMMAND_EXECUTION]: Uses browser-devtools-cli and node-devtools-cli to perform monitoring and debugging actions. These tools are necessary for the skill's stated purpose and are used according to best practices.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 10, 2026, 11:15 AM