council

Pass

Audited by Gen Agent Trust Hub on Apr 6, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill defines a high-level organizational workflow for multi-agent reasoning. It functions as a structured prompt template and does not perform any dangerous actions such as remote code execution, credential harvesting, or unauthorized network communication.
  • [PROMPT_INJECTION]: The skill exhibits an attack surface for indirect prompt injection (Category 8) by design, as it ingests external data to inform decision-making.
  • Ingestion points: Step 2 of the workflow involves gathering codebase-specific files, snippets, issue text, or metrics.
  • Boundary markers: The prompt template in Step 4 uses a Context: label to delimit external data, but does not include explicit instructions for subagents to ignore potentially malicious commands embedded within that data.
  • Capability inventory: The skill's primary output is a text-based recommendation synthesized from subagent responses; it does not directly invoke high-risk system commands or perform file writes (explicitly deferring to external tools like knowledge-ops).
  • Sanitization: The skill does not define specific sanitization or filtering logic for the ingested context.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 6, 2026, 04:00 AM
Security Audit — agent-trust-hub — council