reels-scripting
Warn
Audited by Gen Agent Trust Hub on Apr 28, 2026
Risk Level: MEDIUMCOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill dynamically creates and executes a Node.js script (
analyse-reel.js) within the~/Desktop/Reels/directory to facilitate data scraping and video processing. - [EXTERNAL_DOWNLOADS]: Downloads video files (.mp4) from external Instagram URLs to the local file system during the execution of the generated Node.js script.
- [PROMPT_INJECTION]: Contains instructions aimed at controlling the agent's initial behavior, such as 'CRITICAL: Auto-start on load' and 'do not summarise', which attempt to override default conversational patterns.
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content from external Instagram Reels and Notion links to generate new scripts. It lacks explicit boundary markers or sanitization to prevent malicious instructions within that external data from influencing the agent's output.
- Ingestion points: Step 1 (Notion links and Reel URLs) and Step 2 (Newsletter content).
- Boundary markers: Absent; the skill interpolates untrusted data directly into instructions for the script-writing phase.
- Capability inventory: File writing (
~/Desktop/Reels/), local script execution (Node.js), and network requests (Apify API, Google AI API). - Sanitization: None detected; the skill treats processed transcripts and user inputs as trusted formatting references.
Audit Metadata