improve
Pass
Audited by Gen Agent Trust Hub on Oct 1, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill utilizes shell commands like mkdir and cp to manage the project structure and safeguard original skill files during the experimentation phase.
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process other instruction files, which presents a surface for indirect prompt injection. 1. Ingestion points: Reads local SKILL.md and CLAUDE.md files into context. 2. Boundary markers: The agent performs internal reasoning and logs the evaluation process to a dedicated log file. 3. Capability inventory: File read, write, and copy access within the project directories. 4. Sanitization: External content from target skills is interpolated into the agent's evaluation prompts without specific filtering or escaping mechanisms.
Audit Metadata