manim-idea-to-export
Pass
Audited by Gen Agent Trust Hub on Feb 17, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS] (LOW): The README.md provides installation commands that download the skill from a GitHub repository (
Skheiller/manim-community-deep-skill) not associated with the predefined trusted organizations. While GitHub is a common distribution platform, the source remains unverifiable. - [COMMAND_EXECUTION] (LOW): The skill's primary function involves instructing the agent to run CLI commands like
manim renderandpip install. These are essential for the animation workflow. The skill mitigates risk by advising the use of project-local virtual environments and explicitly prohibiting destructive commands. - [PROMPT_INJECTION] (LOW): The skill has an indirect prompt injection surface (Category 8) because it processes untrusted user input (animation ideas) to generate executable Python code.
- Ingestion points:
SKILL.md(Workflow Step 1) andREADME.md(Example Prompts). - Boundary markers: Absent; user input is treated as instructional context without clear delimiters.
- Capability inventory:
manim rendercalls inSKILL.mdandreferences/render-and-export-playbook.mdinvolve subprocess execution of generated scripts. - Sanitization: Absent; the skill relies on the LLM to filter or properly interpret user input during the code generation phase.
Audit Metadata