phase-self-check
Pass
Audited by Gen Agent Trust Hub on Jun 13, 2026
Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [PROMPT_INJECTION]: The skill defines a process for ingesting untrusted data from user intent and phase observations to guide agent behavior. 1. Ingestion points: User intent and phase observations (SKILL.md). 2. Boundary markers: No delimiters or instructions are present to ignore embedded commands. 3. Capability inventory: The skill uses memory writing and local
perlcommand execution (SKILL.md). 4. Sanitization: No validation or sanitization of external content is specified before interpolation. - [COMMAND_EXECUTION]: The skill suggests using a
perlone-liner (perl -i -pe 's|old|new|g' files...) for bulk file edits. This constitutes a script generation pattern from templates based on the current task state.
Audit Metadata