An MCP server that executes Manim animation code and optionally uploads rendered artifacts to Supabase Storage
Two tools with minimal descriptions and incomplete schemas. execute_manim_code has a 24-char description (below 20-char threshold) and accepts only a single string parameter with no validation, format constraints, or error guidance. cleanup_manim_temp_dir similarly lacks depth. Neither tool documents output structure, error cases, or recovery paths. No parameter validation, no enums, no rate limiting. Tool names are verb-noun but lack clarity on side effects and destructive operations. No tool annotations (destructiveHint, readOnlyHint) despite cleanup being destructive. Missing idempotency guarantees and confirmation patterns for irreversible operations.
Clean up the specified Manim temporary directory after execution.
Execute the Manim code
execute_manim_code description is 24 characters ('Execute the Manim code'), below 20-char threshold and lacks WHAT, WHEN, and prerequisites. LLMs cannot determine when to select this tool vs alternatives.
No input validation or constraints on manim_code parameter. Accepts arbitrary Python strings with no format hints, length limits, or examples. LLMs will pass malformed code without guidance.
cleanup_manim_temp_dir is destructive (WRITE risk) but lacks tool annotation (destructiveHint), confirmation pattern, or dry-run support. No error recovery guidance. Agents cannot safely reason about side effects.
No documented output schemas for either tool. execute_manim_code returns a string but does not specify structure (success/failure, file paths, error details). LLMs cannot extract downstream references.
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 40 | <=2025-11-25 | v2 |
No error classification or recovery guidance. Subprocess failures, file I/O errors, and Supabase upload failures are caught but return raw exception strings. LLMs receive no actionable next steps.