decision-logger

Pass

Audited by Gen Agent Trust Hub on Aug 27, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill framework describes a system architecture that ingests untrusted data from external sources, creating a potential surface for indirect prompt injection.
  • Ingestion points: The "Security & Multi-Agent Hygiene" section identifies web content and external data (e.g., via BrowserOS) as input sources.
  • Boundary markers: The framework includes instructions to treat these inputs as "hostile" and provides guidelines for redacting secrets.
  • Capability inventory: The instructions reference an ExecutionProxy Interface for running terminal commands (e.g., rtk npm test) and a MemoryStore Interface for writing persistent markdown artifacts to the file system (e.g., Obsidian docs/departments/).
  • Sanitization: Guidelines require redacting PII and secrets before sharing context with subagents, providing a layer of data protection during processing.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 27, 2026, 07:35 AM
Security Audit — agent-trust-hub — decision-logger