kuaishou-video

Pass

Audited by Gen Agent Trust Hub on Nov 5, 2026

Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill uses child_process.spawn and spawnSync to execute the ego-browser command-line tool and node scripts. This is the primary mechanism for controlling the automated browser environment.
  • [INDIRECT_PROMPT_INJECTION]: The skill accepts user-supplied video descriptions and comment text. These inputs are typed directly into the web interface of the creator platform. While this represents a standard data ingestion surface, it technically allows for instructions to be passed to the target platform's UI.
  • Ingestion points: Arguments passed to publish.sh (description) and comment.sh (comment text), which are stored in /tmp/kuaishou-video-publish.json and /tmp/kuaishou-video-comment.json.
  • Boundary markers: None identified for the description/comment text fields; they are inserted as-is using document.execCommand('insertText', ...).
  • Capability inventory: File system read (size/existence), Browser automation (clicking, typing, uploading), and subprocess execution.
  • Sanitization: Basic validation checks for empty strings and file existence are performed before processing.
  • [DYNAMIC_EXECUTION]: The skill dynamically generates a temporary JavaScript entry file (entry.mjs) to bridge the execution environment between the local shell and the ego-browser tool. It also uses page.evaluate to execute scripts within the context of the browser page, which is standard for web automation tasks.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 5, 2026, 06:58 AM