exercise-asb-skill
Pass
Audited by Gen Agent Trust Hub on Jul 22, 2026
Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [PROMPT_INJECTION]: The skill is designed to ingest and process content from external SKILL.md files and simulated user inputs, which presents a surface for indirect prompt injection. A malicious skill file could contain instructions intended to override the testing agent's logic.
- Ingestion points: Reads file content from target skills specified by name or path in Step 1.
- Boundary markers: Instructions define a 'quarantine' where the agent must only use information from the target file and recommends structural isolation using subagents.
- Capability inventory: The skill utilizes Read, Glob, Grep, and Edit tools to interact with the file system.
- Sanitization: No specific sanitization or escaping of the target file's content is described prior to its use in role-play scenarios.
- [COMMAND_EXECUTION]: The skill utilizes the Edit tool to perform file system modifications, specifically to apply patches to the SKILL.md files it is exercising. This allows the agent to write content to the local environment to fix detected failures.
Audit Metadata