crash-test
Pass
Audited by Gen Agent Trust Hub on Oct 23, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user input from text or files, which serves as a potential surface for indirect prompt injection attacks where malicious data could influence sub-agent behavior.\n
- Ingestion points: Business plans are ingested from user-provided text or file paths (
SKILL.md).\n - Boundary markers: There are no defined boundary markers or XML tags to encapsulate the user-provided data within the analysis prompts.\n
- Capability inventory: The skill utilizes sub-agent orchestration and file system write access (
SKILL.md).\n - Sanitization: No sanitization or safety-filtering is specified for the ingested content before it is distributed to the "crash dummy" agents.\n- [COMMAND_EXECUTION]: The skill automates the creation of directories and files on the local system.\n
- Evidence: Instructs the agent to create a
.crash-test/directory and write markdown report files (SKILL.md).\n - Mitigation: The skill includes a 'least privilege' directive, explicitly instructing the agent to 'Touch no other files,' which mitigates the risk of unauthorized file system modifications.
Audit Metadata