trunative-web
Pass
Audited by Gen Agent Trust Hub on Sep 17, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The AI agent is instructed to read and follow project-specific briefs and identity documentation which could serve as a vector for indirect injection attacks.
- Ingestion points: The agent reads from PRODUCT.md, DESIGN.md, and STACK.md project files, as well as screen briefs in the .trunative/screens/ directory.
- Boundary markers: No explicit boundaries or markers are defined to distinguish between internal agent instructions and untrusted project data.
- Capability inventory: The skill can execute shell commands (via npx), capture screenshots using headless Chrome, and write report files to the local filesystem.
- Sanitization: No sanitization of the markdown content within project files is implemented.
- [COMMAND_EXECUTION]: The skill workflow involves executing shell commands to perform analysis and manage its lifecycle.
- Evidence: Invokes 'npx trunative doctor', 'npx trunative install', and 'npx @google/design.md' in flow/init.md.
- Evidence: Uses 'chrome --headless' for screenshot generation and 'npx trunative detect' for code scanning in flow/review.md.
- [EXTERNAL_DOWNLOADS]: The skill depends on external Node.js packages to provide its core analysis and validation functionality.
- Evidence: References 'trunative' and '@google/design.md' which are typically managed via npm/npx.
Audit Metadata