code-reviewer

Pass

Audited by Gen Agent Trust Hub on Aug 27, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill identifies and attempts to manage an attack surface involving the ingestion of untrusted external data.
  • Ingestion points: Web content and external data are ingested via the BrowserOS interface as described in the 'Security & Multi-Agent Hygiene' section.
  • Boundary markers: The prompt lacks specific technical delimiters (e.g., XML tags or unique markers) to isolate untrusted content, though it instructs the agent to treat such inputs as 'hostile'.
  • Capability inventory: The agent has the power to execute terminal commands through the ExecutionProxy Interface (rtk prefix) and write to persistent storage via the MemoryStore Interface (docs/departments/).
  • Sanitization: The instructions explicitly mandate the redaction of secrets and PII (Personally Identifiable Information) before sharing context with other agents.
  • [COMMAND_EXECUTION]: The skill references an ExecutionProxy Interface used to perform terminal actions (e.g., rtk npm test), which grants the agent the capability to execute shell commands within its runtime environment.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 27, 2026, 07:35 AM
Security Audit — agent-trust-hub — code-reviewer