pandastudio

Pass

Audited by Gen Agent Trust Hub on Nov 1, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted text data through the transcription of user-uploaded videos and external media downloads. This content is then interpolated into prompts for a local LLM to generate titles, descriptions, and timestamps. This process presents a risk where malicious instructions hidden in the transcript could influence the agent's behavior.\n
  • Ingestion points: Video transcripts processed by transcript.get, and content from external URLs fetched via media.download-url or media.import.\n
  • Boundary markers: Not explicitly enforced in the instructions provided for LLM metadata generation verbs like llm.generate-title.\n
  • Capability inventory: Extensive file system operations (project management), network access via connectors, and social media publishing capabilities (export.publish-social, export.publish-youtube).\n
  • Sanitization: No explicit content filtering or sanitization of the transcript text is documented before it is passed to the LLM.\n- [EXTERNAL_DOWNLOADS]: The skill automates the download of essential media binaries and AI models upon first use. This includes FFmpeg for media encoding, yt-dlp for online video downloads, and various local models for transcription (Whisper) and narration (Kokoro).\n
  • Evidence: Found in system.install-media-engine, media.download-url, and system.download-whisper-model.\n
  • Note: These are well-known technology tools and models necessary for the skill's primary media-editing purpose. The skill includes checksum verification for specific tools like yt-dlp.\n- [COMMAND_EXECUTION]: The skill relies on the execution of shell commands through the pandastudio CLI binary to perform its editing and system management tasks.\n
  • Evidence: Found throughout SKILL.md and reference/commands.md.\n
  • Description: The skill provides an authoritative playbook for using CLI verbs to manipulate project state, manage files, and invoke local processing tools. This behavior is consistent with its role as an interface for a local video editor.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 1, 2026, 09:11 PM