arabic-captions

Pass

Audited by Gen Agent Trust Hub on Nov 9, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill provides practical guidance for addressing common layout issues in Arabic subtitles, such as character shaping and punctuation placement.
  • [OBFUSCATION]: The instructions utilize the Right-to-Left Mark (U+200F) and Byte Order Mark (U+FEFF). Although these are invisible control characters often associated with obfuscation, their usage here is contextually appropriate for ensuring correct RTL text rendering and file interoperability in video editing suites like Premiere Pro.
  • [COMMAND_EXECUTION]: Provides a standard ffmpeg command template for burning subtitles. The command uses established filters (subtitles, libass) and does not involve network activity or suspicious parameters.
  • [INDIRECT_PROMPT_INJECTION]: The skill defines a workflow for processing transcripts, representing an untrusted data ingestion surface.
  • Ingestion points: Transcript content used to generate .srt and .ass files.
  • Boundary markers: None explicitly defined in the provided snippets.
  • Capability inventory: ffmpeg subprocess execution for video encoding.
  • Sanitization: Uses basic whitespace stripping (line.strip()) in formatting logic. The risk is considered minimal as the logic is formatting-centric rather than decision-making.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 9, 2026, 12:32 PM