test-report-review
Pass
Audited by Gen Agent Trust Hub on Oct 14, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill is designed as a security-conscious auditor for test reports. It explicitly forbids the agent from accepting report claims at face value without supporting raw evidence (logs, defects) and includes strict invariants to prevent project stakeholders from overriding quality facts or technical failure results. The workflow mandates that human authority is preserved by keeping the final decision state as pending.
- [INDIRECT_PROMPT_INJECTION]: The skill has an input surface for processing untrusted reports, but no exploitable capabilities were found.
- Ingestion points: Processes user-supplied test reports, execution evidence (logs/screenshots), defect lists, and external role reports (QA, Tech, Product) via the agent prompt.
- Boundary markers: Relies on a mandatory versioned source inventory and an evidence-consistency matrix to isolate report claims from raw evidence; it does not utilize specific character-based delimiters like triple-dashes for input separation.
- Capability inventory: The skill instructions and agent manifest do not request or utilize network access, file system write operations, or arbitrary command execution capabilities.
- Sanitization: Implements mandatory semantic reconciliation where every report claim must be validated against independent raw source evidence (e.g., matching a 'pass' claim to a specific log entry and build hash) before an AI-assisted recommendation is issued.
Audit Metadata