talking-head-guide

Pass

Audited by Gen Agent Trust Hub on Jul 29, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: Instructional Integrity: The content consists entirely of legitimate operational guidelines for video editing workflows, including A-roll cleanup, B-roll placement, and caption management.
  • [SAFE]: Platform Tooling: The skill references internal platform tools like read_script, apply_script, and multicam_sync for their intended purposes without any attempt to bypass restrictions or escalate privileges.
  • [SAFE]: No Malicious Patterns: There is no evidence of prompt injection, data exfiltration, obfuscation, or unauthorized network communication. All workflows are focused on media processing within the designated environment.
  • [SAFE]: Indirect Prompt Injection Surface: The skill interacts with user-provided transcripts, but it establishes clear structural boundaries using segment markers (e.g., [sN]) and specific editing syntax, which mitigates the risk of the agent misinterpreting data as instructions.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 29, 2026, 03:02 PM
Security Audit — agent-trust-hub — talking-head-guide