sprite-forge
Fail
Audited by Gen Agent Trust Hub on Nov 1, 2026
Risk Level: CRITICALREMOTE_CODE_EXECUTIONEXTERNAL_DOWNLOADSDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [REMOTE_CODE_EXECUTION]: The skill repeatedly instructs the user to install the 'genmedia' CLI by piping a remote script directly into bash, which is an extremely dangerous pattern for executing unverified code from an untrusted source. Evidence: 'curl https://genmedia.sh/install -fsS | bash' is explicitly recommended in SKILL.md, README.md, scripts/gen_image.py, and references/provider-matrix.md.
- [EXTERNAL_DOWNLOADS]: The documentation suggests downloading and adding components from an untrusted third-party domain (asciistudio.space), increasing the risk of supply chain attacks. Evidence: 'npx shadcn@latest add https://asciistudio.space/r/.json' in references/ascii-animation-components.md.
- [DYNAMIC_EXECUTION]: Several scripts, notably the pipeline runner, dynamically construct and execute shell commands using subprocess calls, creating a broad attack surface for command injection if configuration files or variables are compromised. Evidence: scripts/run_pipeline.py (line 218), scripts/turnaround.py (line 39), scripts/gen_image.py (lines 69, 78, 102), and scripts/video_to_spritesheet.py (line 42).
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted user prompts and image data to generate game assets and orchestrate multi-stage pipelines, lacking sufficient boundary markers to prevent instructions embedded in that data from influencing the agent's behavior. Ingestion points: User-provided prompts and reference images in turnaround.py, gen_image.py, and video_to_spritesheet.py. Boundary markers: The skill does not use delimiters or instructions to ignore embedded commands. Capability inventory: The skill has extensive capabilities including subprocess.run across multiple scripts, file writing in run_pipeline.py, and network operations. Sanitization: No evidence of escaping or validation of external content.
Recommendations
- HIGH: Downloads and executes remote code from: https://genmedia.sh/install - DO NOT USE without thorough review
- AI detected serious security threats
Audit Metadata