test-a-skill
Pass
Audited by Gen Agent Trust Hub on Sep 21, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze external skill files provided as the argument. This creates an attack surface where a malicious target skill could contain instructions intended to manipulate the testing agent's evaluation or hide defects. 1. Ingestion points: The skill takes a file path or identifier as a command argument (/test-a-skill ) in SKILL.md. 2. Boundary markers: No specific delimiters or "ignore embedded instructions" warnings are provided to separate the target skill's content from the testing agent's system prompt. 3. Capability inventory: The instructions suggest giving testing agents the tools they need, including file system access to verify claims against source code. 4. Sanitization: No instructions for sanitizing or validating the content of the target skill are present.
- [COMMAND_EXECUTION]: The instructions direct the agent to execute shell commands such as 'claude -p' to spawn headless agent sessions for parallel testing, which involves direct invocation of external CLI tools.
Audit Metadata