skills/hktitan/skills/khe-package/Gen Agent Trust Hub

khe-package

Pass

Audited by Gen Agent Trust Hub on Nov 17, 2026

Risk Level: SAFECOMMAND_EXECUTIONDYNAMIC_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill relies on local command execution via node:child_process to perform its core functions. It spawns ffmpeg for video encoding and ffprobe for media metadata verification. It also executes its own maintenance and rendering scripts (render.mjs, package-check.mjs) as part of the documented workflow.
  • [DYNAMIC_EXECUTION]: The renderer script (render.mjs) utilizes dynamic imports to load the playwright-core library from variable paths. Additionally, it uses page.evaluate() to inject and execute JavaScript within the Playwright browser context to manipulate animation timing and synchronize the 'paused clock' rendering mechanism.
  • [EXTERNAL_DOWNLOADS]: Documentation in the README and reference files instructs users to install playwright-core and system utilities like ffmpeg and Google Chrome. These are standard requirements for the skill's stated purpose of automated video rendering.
  • [INDIRECT_PROMPT_INJECTION]: The skill possesses an indirect prompt injection surface as it ingests and processes untrusted local data, specifically HTML explainer pages and JSON shot lists. While the agent primarily uses these for rendering and validation, malicious instructions embedded in these source files could potentially influence the agent's interpretation during the verification and review steps.
  • Ingestion points: examples/explainer/index.html, examples/live/stage.html, and user-provided HTML/JSON source files.
  • Boundary markers: The skill uses a structured PACKAGE.md plan to delimit metadata, though it lacks explicit 'ignore instructions' markers for the content of the HTML sources themselves.
  • Capability inventory: File system access (reading source files, writing videos), command execution (ffmpeg, node), and browser-based script execution.
  • Sanitization: No explicit sanitization of external HTML or JSON content is performed before processing or rendering.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 17, 2026, 06:12 AM