requirements-analysis-plus
Warn
Audited by Gen Agent Trust Hub on Oct 14, 2026
Risk Level: MEDIUMINDIRECT_PROMPT_INJECTIONPRIVILEGE_ESCALATION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user-supplied documents in multiple formats (Word, Excel, HTML, JSON, Markdown), creating an indirect prompt injection surface.
- Ingestion points: Parsing functions in scripts/common_parser.py and scripts/run_analysis.py.
- Boundary markers: The main prompt in prompts/requirements-analysis-plus.md requires separating confirmed facts from inferences and mandates item-by-item source attribution.
- Capability inventory: scripts/run_analysis.py performs local file system reading and writes analysis reports to the disk.
- Sanitization: scripts/evidence_scaffold.py implements SHA-256 hashing to track source integrity, and script outputs are explicitly labeled as NOT_ASSESSED to ensure agent-level verification.
- [PRIVILEGE_ESCALATION]: The README.md file contains installation instructions for Windows that use the -ExecutionPolicy Bypass flag with PowerShell. Evidence: powershell -ExecutionPolicy Bypass -File .\scripts\install-skills-windows.ps1. This flag allows the execution of local scripts by bypassing default security restrictions, which is a privilege escalation vector that requires user review of the script content.
Audit Metadata