yetong-fitness-xhs-copy
Pass
Audited by Gen Agent Trust Hub on Oct 3, 2026
Risk Level: SAFECOMMAND_EXECUTIONDATA_EXFILTRATIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: No security issues or malicious patterns were detected. The skill implements strong defensive measures, such as prohibiting the creation of fake testimonials and requiring explicit authorization for using member stories.
- [COMMAND_EXECUTION]: The skill utilizes a local script (scripts/check_profile.py) to validate the structure of profile files. This script is limited to structural parsing using standard Python libraries and does not perform risky operations or execute external code.
- [DATA_EXFILTRATION]: Instructions strictly limit file access to specific, user-provided character profiles (person-dna.md) and prohibit searching the host system for additional private data. No network functionality is present to exfiltrate information.
- [INDIRECT_PROMPT_INJECTION]: The skill processes external profile data but includes robust boundary markers and verification steps. 1. Ingestion points: User-supplied person-dna.md file and copywriting topics. 2. Boundary markers: The references/voice-and-evidence.md file defines strict usage tiers (confirmed vs. internal-only) and requires verification of facts before they are included in copy. 3. Capability inventory: Limited to text generation and a local structural validation script; no file-writing or network capabilities are present. 4. Sanitization: While the validation script checks file structure, the skill's instructions require the agent to manually verify the semantic context and authorization of all ingested facts.
Audit Metadata