yetong-dna
Pass
Audited by Gen Agent Trust Hub on Oct 17, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill runs a local script,
scripts/validate_profile.py, to ensure the generated profile follows the correct format. This script only uses standard Python tools and does not access the internet. - [INDIRECT_PROMPT_INJECTION]: The skill collects user interview answers to build a profile. While this involves handling user-provided text, the skill includes several safety steps, such as labeling profiles as 'draft' and requiring user confirmation before use.
- Ingestion points: User input during interviews as described in
SKILL.md. - Boundary markers: Present; uses status levels like 'draft' and 'active' with user confirmation.
- Capability inventory: Writes a local markdown file and runs the local validation script.
- Sanitization: Relies on AI categorization and user review.
- [SAFE]: No malicious code or exfiltration risks were found. The skill follows best practices for data integrity and privacy.
Audit Metadata