architect
Pass
Audited by Gen Agent Trust Hub on Aug 27, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill defines a processing surface for untrusted external data which could contain malicious instructions.
- Ingestion points: Web content and external data processed via the 'BrowserOS' tool as described in SKILL.md.
- Boundary markers: The skill lacks technical delimiters (e.g., XML tags or specific markdown blocks) to isolate untrusted data from the primary instruction stream.
- Capability inventory: The skill utilizes file writing through the 'MemoryStore' interface and shell command execution via the 'ExecutionProxy'.
- Sanitization: The instructions include a manual requirement to redact secrets and PII before sharing context with subagents, which serves as a basic manual filter.
- [COMMAND_EXECUTION]: The skill mandates the use of an 'ExecutionProxy' interface (using the 'rtk' prefix) to perform terminal actions like 'npm test'. While this is part of a deterministic testing protocol (TDD), it explicitly grants and structures shell execution capabilities within the agent's workflow.
Audit Metadata