governance
Pass
Audited by Gen Agent Trust Hub on Mar 7, 2026
Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [PROMPT_INJECTION]: The skill performs persistent instruction injection by modifying project context files (
CLAUDE.mdor.cursor/rules/swain-governance.mdc). These injected instructions are designed to override the agent's default operational logic: - They mandate that the agent must "ALWAYS" use the
spec-managementskill for documentation and "do not improvise artifact creation." - They explicitly forbid the use of "built-in agent todos" or "markdown TODOs," forcing the agent to use the
execution-trackingskill and an external tool calledbd (beads)instead. - The instructions use high-authority markers like "MANDATORY" and "do not edit this block manually" to ensure compliance and persistence.
- [COMMAND_EXECUTION]: The skill uses the
bashtool to executegrep, checking for the existence of governance strings in the project root to determine if installation is required.
Audit Metadata