refactor-cleaner

Pass

Audited by Gen Agent Trust Hub on Aug 27, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill instructions define an attack surface where untrusted data could influence agent behavior during high-privilege operations.
  • Ingestion points: The 'Security & Multi-Agent Hygiene' section explicitly mentions 'BrowserOS' as a source of hostile external data.
  • Boundary markers: The instructions do not define specific delimiters or framing techniques to isolate external data from system instructions.
  • Capability inventory: The protocols define the use of an 'ExecutionProxy Interface' (defaulting to the rtk prefix) for terminal actions such as rtk npm test, indicating the agent environment has shell execution capabilities.
  • Sanitization: While the protocols mandate the redaction of secrets and PII, they do not specify technical sanitization or validation methods to prevent malicious instructions embedded in external web content from being executed via the defined command interfaces.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 27, 2026, 07:35 AM
Security Audit — agent-trust-hub — refactor-cleaner