acceptance-criteria-review
Pass
Audited by Gen Agent Trust Hub on Oct 31, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to process external, potentially untrusted data such as user requirements and business rules to generate its analysis.
- Ingestion points: Textual inputs including user stories, requirements, and historical results are processed as described in prompts/acceptance-criteria-review.md.
- Boundary markers: The skill includes instructions to audit input completeness and requires the agent to separate facts from assumptions and identify open questions.
- Capability inventory: The skill is primarily focused on generating markdown documentation; it includes local script execution for evaluation purposes in the evals/ directory.
- Sanitization: The skill relies on natural language instructions to avoid inventing system behavior and to mark unsupported conclusions as hypotheses.- [COMMAND_EXECUTION]: The skill package includes local shell and Python scripts used for deterministic validation of analysis results within the provided evaluation suite.
- Evidence: evals/fixtures/check-domain-decisions.sh executes evals/fixtures/check_domain_decisions.py to verify formatted table data.
- Context: These scripts are static, perform local parsing/validation, and are used to ensure the quality of the skill's outputs against a predefined oracle.
Audit Metadata