zj-edit-article
Pass
Audited by Gen Agent Trust Hub on Sep 4, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: No malicious instructions, obfuscation, or data exfiltration mechanisms were identified. The skill's scope is strictly limited to text editing and restructuring.
- [INDIRECT_PROMPT_INJECTION]: The skill ingests and transforms article drafts provided as user input (SKILL.md), which constitutes a potential ingestion point for untrusted data. Analysis of the execution context confirms a total absence of capabilities for network operations, file system modifications, or shell command execution. Due to this restricted capability inventory, and despite the absence of explicit boundary markers or input sanitization, any adversarial instructions embedded in the article text cannot be actioned, rendered the surface non-exploitable.
Audit Metadata