crucible-meta-governance

Warn

Audited by Gen Agent Trust Hub on Apr 20, 2026

Risk Level: MEDIUMCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [DYNAMIC_EXECUTION]: The skill includes 'Self-Evolving Skill' instructions that direct the agent to autonomously update 'SKILL.md' and 'references/evolution-log.md' if it determines that patterns are misleading or require improvement. This establishes a self-modifying instruction pattern.
  • [COMMAND_EXECUTION]: The skill provides specific shell command snippets, such as 'mkdir' and 'cp', and instructs the agent to perform file management tasks to move artifacts from temporary locations to persistent storage.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data, including session examples and outputs from other agents, which could influence meta-governance decisions or trigger autonomous updates to the skill's instructions.
  • [INDIRECT_PROMPT_INJECTION]: Ingestion points: Untrusted data is read using the 'Read', 'Grep', and 'Glob' tools specified in the frontmatter.
  • [INDIRECT_PROMPT_INJECTION]: Boundary markers: The skill does not define delimiters or provide instructions for the agent to disregard commands or instructions embedded within the processed data.
  • [INDIRECT_PROMPT_INJECTION]: Capability inventory: Across the instruction set, the agent is granted the ability to modify its own instruction files and execute file system management operations.
  • [INDIRECT_PROMPT_INJECTION]: Sanitization: There are no instructions for validating, escaping, or sanitizing external content before it is used to inform decisions or update the skill's logic.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Apr 20, 2026, 10:09 PM