spec-driven-workflow
Pass
Audited by Gen Agent Trust Hub on Aug 21, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill implements engineering best practices using local scripts and documentation. Technical analysis of the provided Python scripts (
spec_validator.py,spec_generator.py, andtest_extractor.py) confirms they rely exclusively on the Python standard library (including argparse, json, re, and pathlib) for all operations. The skill does not perform network requests, access unauthorized file system paths, or utilize dangerous functions such as eval() or exec().\n- [INDIRECT_PROMPT_INJECTION]: The workflow involves processing user-provided or agent-generated markdown specification files, which creates a potential surface for indirect prompt injection. \n - Ingestion points: The
spec.mdfile is ingested by the validator and extractor scripts to produce scores and code stubs.\n - Boundary markers: The scripts utilize explicit markdown section headers (e.g.,
## Functional Requirements,### AC-N:) to identify and scope processed content.\n - Capability inventory: The skill enables local file reading, script execution for validation, and template-based test generation.\n
- Sanitization: The
test_extractor.pyscript performs basic alphanumeric sanitization for function names, though requirement text is preserved in comments. The overall risk is mitigated by the structured workflow which mandates human-led Phase 3 (Validation) and Phase 6 (Self-Review) steps.\n- [DYNAMIC_EXECUTION]: Thetest_extractor.pyutility generates test stubs for various frameworks (pytest, jest, go-test) by extracting data from markdown files. This is a core feature of the workflow and employs static text templates for generation rather than executing the extracted content, resulting in a safe implementation.
Audit Metadata