cro
Pass
Audited by Gen Agent Trust Hub on Sep 25, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill defines a workflow that ingests untrusted external data, creating a potential surface for indirect prompt injection.
- Ingestion points: External data and web content accessed via 'BrowserOS'.
- Boundary markers: The instructions explicitly state that external data must be treated as hostile.
- Capability inventory: The agent uses an 'ExecutionProxy' for terminal commands (e.g., 'rtk npm test') and a 'MemoryStore' for writing to markdown files in 'docs/'.
- Sanitization: The skill mandates the redaction of secrets and PII before sharing context with subagents.
- [SAFE]: The skill uses vendor-specific tools and references ('galyarder-framework') that are consistent with the author's identity. The use of standard development tools like 'npm' and 'Linear' for project management and testing is typical for software engineering tasks.
Audit Metadata