generating-test-data
Pass
Audited by Gen Agent Trust Hub on Aug 21, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [DYNAMIC_EXECUTION]: The script
scripts/generate_data.pyincludes agenerate_scriptmethod that writes arbitrary content to a shell script (.sh) and marks it as executable usingchmod 0o755. This allows for the creation of executable files from unvalidated input. - [INDIRECT_PROMPT_INJECTION]: The
SKILL.mdinstructions direct the agent to generate 'Adversarial inputs' such as SQL injection strings, XSS payloads, and Unicode edge cases. This introduces a risk of indirect prompt injection where generated values could influence subsequent agent actions or system behavior if not properly handled. - Ingestion points: Data models, schemas, and type definitions read by the skill.
- Boundary markers: No delimiters or ignore instructions identified for the generated adversarial content.
- Capability inventory: Access to
WriteandEdittools, plus a script for creating executable files. - Sanitization: No sanitization is performed on the data written to generated files.
Audit Metadata