chief-of-staff
Pass
Audited by Gen Agent Trust Hub on Sep 25, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [SAFE]: The skill acts as a set of governing protocols for agent behavior, focusing on operational transparency and technical integrity through defined modes like BUILD and INCIDENT.
- [INDIRECT_PROMPT_INJECTION]: The skill acknowledges the risks of external data ingestion. 1. Ingestion points: Web content and external data via BrowserOS. 2. Boundary markers: Explicit instructions to treat external data as hostile. 3. Capability inventory: Command execution through ExecutionProxy and markdown artifact storage via MemoryStore. 4. Sanitization: Mandated redaction of secrets and PII before context is shared with subagents.
- [COMMAND_EXECUTION]: The framework utilizes an 'ExecutionProxy' (rtk prefix) for running terminal commands such as 'npm test'. This design is intended to minimize computational overhead and enforce deterministic gating for code changes.
Audit Metadata