squad-product
Pass
Audited by Gen Agent Trust Hub on Oct 23, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data (requests, documents, and screenshots) and implements robust mitigations.
- Ingestion points: User requests, linked issues, pasted documents, and screenshots as specified in
SKILL.md. - Boundary markers: The skill includes explicit instructions to treat all external inputs as untrusted data and ignore any instructions embedded within them.
- Capability inventory: The agent is permitted to write plan documents to the file system but is strictly forbidden from executing code or running tests.
- Sanitization: Instructions include redacting secrets and personal data before processing.
- [EXTERNAL_DOWNLOADS]: The skill references several official regulatory and standards websites for verification purposes.
- Evidence: Links to W3C, Apple, Google, and official government policy sites (GDPR, PDPA, etc.) in
references/official-sources.md. These are trusted informational sources and do not involve executable code. - [SAFE]: The skill enforces a limited scope of operation, explicitly stating it writes no code and advances no gates, focusing purely on requirement analysis and documentation.
Audit Metadata