council

Pass

Audited by Gen Agent Trust Hub on Jun 20, 2026

Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [PROMPT_INJECTION]: The skill implements a workflow where user-provided topics are interpolated directly into instruction templates for simulated sub-agent personas. This creates a surface for indirect prompt injection where a maliciously crafted query could attempt to influence persona behavior.
  • Ingestion points: User question/topic provided via $ARGUMENTS; local project files (CLAUDE.md and .claude/rules/).
  • Boundary markers: User-provided text is wrapped in double quotes within the persona prompt template.
  • Capability inventory: The skill performs local file writes to record analysis results (e.g., council-[topic-slug].md).
  • Sanitization: No explicit sanitization or validation of the input strings is performed before they are embedded in sub-agent prompts.
  • [COMMAND_EXECUTION]: The skill performs file system operations by reading project-level configuration files and writing synthesized reports to the local directory. These actions are transparent and align with the skill's stated purpose of providing contextual decision support.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 20, 2026, 06:11 PM
Security Audit — agent-trust-hub — council