charles-claude-editing

Pass

Audited by Gen Agent Trust Hub on Nov 17, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill possesses a vulnerability surface for indirect prompt injection due to the processing of untrusted external content.
  • Ingestion points: Processes output from transcription services, reference videos provided by users, and media assets (images/videos) retrieved from the open web during asset discovery.
  • Boundary markers: There are no explicit instructions to the agent to treat external content as untrusted data or to ignore embedded instructions within that data.
  • Capability inventory: The agent is authorized to perform file system writes (specifically saving to Apple Photos), conduct web searches, and invoke other functional skills (charles-arcade-assets and charles-explainers-assets).
  • Sanitization: The instructions lack requirements for sanitizing or validating the content of ingested external files, focusing only on output filtering for TikTok safety guidelines.
  • [EXTERNAL_DOWNLOADS]: The skill facilitates the retrieval of external resources.
  • Automatically searches for and downloads images and videos from the web to be used as b-roll or overlays.
  • Depends on external modules and assets managed via a remote skill repository hosted on GitHub.
  • [COMMAND_EXECUTION]: The skill performs intensive media processing which involves the execution of underlying system tools.
  • Requires the agent to perform complex video editing tasks such as cutting, splicing, and transcoding (H.264, AAC) based on high-level instructions.
  • Explicitly instructs the agent to save finished reels to the Apple Photos application, implying the use of system-level commands or scripts to interact with local databases/applications.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 17, 2026, 02:17 PM