viral-game

Pass

Audited by Gen Agent Trust Hub on Nov 11, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONDYNAMIC_EXECUTIONEXTERNAL_DOWNLOADSDATA_EXFILTRATION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted third-party content from tweets to generate game concepts. Although it includes an explicit 'Content Boundary' policy in tweet-pipeline.md to prevent interpreting tweet text as instructions, the skill's broad capabilities (file writes, shell execution, network access) make this a potential attack surface if the boundaries are bypassed.
  • [COMMAND_EXECUTION]: The orchestrator performs numerous shell operations, including npm install, curl, ffmpeg, and the execution of external shell scripts for deployment (publish.sh) and video processing (convert-highfps.sh). These operations are necessary for the skill's functionality but involve executing multiple subprocesses.
  • [DYNAMIC_EXECUTION]: The pipeline relies on 'Task' subagents to autonomously generate, modify, and verify game code. This involves the dynamic creation and execution of JavaScript and shell scripts based on AI-interpreted concepts and autofix logic.
  • [EXTERNAL_DOWNLOADS]: The skill fetches data and assets from external services including fxtwitter.com, Meshy AI, and World Labs. It also instructs the user to install external agent skills using npx skills add, which is handled with explicit user consent.
  • [DATA_EXFILTRATION]: The skill transmits user-provided email addresses to the here.now API for account setup and sends game metadata to api.play.fun for registration. These network operations are functional requirements of the deployment and monetization features.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 11, 2026, 08:10 AM