behavioral-modes
Pass
Audited by Gen Agent Trust Hub on Feb 16, 2026
Risk Level: LOW
Full Analysis
- [PROMPT_INJECTION] (SAFE): The skill uses instructional markers to define stylistic personas but does not attempt to bypass safety guidelines or override core system prompts maliciously.
- [DATA_EXFILTRATION] (SAFE): No network operations or external communication patterns were found. Tools are restricted to local read-only operations.
- [REMOTE_CODE_EXECUTION] (SAFE): There are no patterns involving the download or execution of remote scripts or packages.
- [COMMAND_EXECUTION] (SAFE): The frontmatter explicitly limits tools to Read, Glob, and Grep, which do not permit arbitrary system command execution or file modification.
- [INDIRECT_PROMPT_INJECTION] (LOW): 1. Ingestion points: DEBUG and REVIEW modes process external error logs and code. 2. Boundary markers: None specified in the templates. 3. Capability inventory: Tools are limited to Read, Glob, and Grep (read-only). 4. Sanitization: No sanitization is specified. The risk is categorized as LOW because the skill lacks the write or execute capabilities required for a successful injection-based attack to impact the system.
Audit Metadata