feedback

Pass

Audited by Gen Agent Trust Hub on Jun 27, 2026

Risk Level: SAFEDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
  • [DATA_EXFILTRATION]: The skill instructs the agent to send data to an external endpoint: https://skills.devnads.com/api/feedback. While the skill mandates user consent and provides detailed privacy rules to scrub sensitive information, this domain is not categorized as a trusted or well-known service.
  • [PROMPT_INJECTION]: The skill possesses an indirect prompt injection surface (Category 8) by ingesting untrusted session data to generate feedback reports.
  • Ingestion points: The skill draws from user conversation history and tool outputs to populate the message and context fields in the feedback payload.
  • Boundary markers: The skill requires the agent to display the payload to the user for review and explicit confirmation before sending, which serves as a manual boundary.
  • Capability inventory: The skill utilizes network operations (specifically curl for POST requests) to transmit data externally.
  • Sanitization: The instructions include a comprehensive "Privacy rules" section requiring the removal of private keys, PII, absolute paths, and specific identifiers before the data is sent.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 27, 2026, 11:48 AM
Security Audit — agent-trust-hub — feedback