research

Pass

Audited by Gen Agent Trust Hub on Aug 29, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill describes a workflow that involves ingesting and processing untrusted external data sources, including codebase surveys, CI logs, and large diffs or pull requests. This workflow creates a surface for indirect prompt injection attacks where malicious instructions hidden in these sources could potentially influence the subagent's distilled output or the orchestrator's subsequent actions.
  • Ingestion points: External data sources like files, logs, and diffs as specified in SKILL.md.
  • Boundary markers: The instructions do not prescribe the use of delimiters or 'ignore embedded instructions' prompts when passing data to subagents.
  • Capability inventory: Utilization of subagent spawning with specific model tiers for search and distillation tasks.
  • Sanitization: No mention of sanitizing or escaping the external content before it is processed by the agent system.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 29, 2026, 08:40 PM
Security Audit — agent-trust-hub — research