freeze
Pass
Audited by Gen Agent Trust Hub on Sep 6, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes decision records which are external files that could contain malicious instructions.
- Ingestion points: The workflow in SKILL.md requires reading a decision record to validate its status before applying a freeze.
- Boundary markers: The instructions do not define delimiters or specific 'ignore' rules to prevent the agent from following instructions embedded within the decision records.
- Capability inventory: The skill performs file system write operations, including updating the decision record itself and writing to a central index file located at
~/.claude/freezes/active.md. - Sanitization: There is no evidence of content sanitization or validation for the data ingested from the decision records.
Audit Metadata