regression-scope-analysis
Pass
Audited by Gen Agent Trust Hub on Oct 31, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to process untrusted external data such as release scopes, change lists, and stakeholder concerns to generate regression plans. This creates a surface for indirect prompt injection.
- Ingestion points: Data enters via the user prompt and processed 'real materials' referenced in
prompts/regression-scope-analysis.md. - Boundary markers: The instructions mandate separating 'confirmed facts', 'working assumptions', and 'open questions', and use markdown headers for output structuring.
- Capability inventory: The skill performs analysis and produces text artifacts; it does not have tools for file system writes or network operations.
- Sanitization: The skill instructions include an 'Input Audit' step to evaluate input completeness and credibility.
- [SAFE]: The execution scripts provided in the
evals/fixtures/directory are standard testing utilities. The Python script (check_domain_decisions.py) performs deterministic parsing and validation of markdown tables against a static oracle without performing any network requests or sensitive file access.
Audit Metadata