instagram-carousel

Pass

Audited by Gen Agent Trust Hub on Sep 15, 2026

Risk Level: SAFECOMMAND_EXECUTIONDATA_EXFILTRATIONINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill instructs the agent to automate the installation of the Playwright library and the Chromium browser using pip3 install playwright and python3 -m playwright install chromium if they are not already present.
  • [DATA_EXPOSURE]: The skill reads local file system content based on user-provided paths (using Path.read_bytes()) to encode them as base64 URIs for embedding in HTML. It lacks directory traversal protections or validation to ensure the files are images, which could be exploited to read sensitive system configuration or credential files if the user is prompted with a malicious path.
  • [INDIRECT_PROMPT_INJECTION]: The skill ingests content from user-provided URLs and brand copy to generate HTML. This content is then rendered in a headless browser (Playwright) to produce the final images. This architecture creates a surface for indirect prompt injection where malicious external content could attempt to execute JavaScript (XSS) or perform actions within the context of the rendering browser.
  • [DYNAMIC_EXECUTION]: The skill generates and writes both a Python script and HTML content to the local filesystem at runtime to facilitate the carousel export and screenshot process.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 15, 2026, 03:48 AM