MCP server for Adobe application automation via Socket.IO proxy and plugin integration
This server has critical structural issues that prevent reliable agent use. Of 7 tools listed, only 2 have visible input schemas in the provided source code (createDocument, executeExtendScript). The remaining 5 tools (getActiveDocumentSettings, getProjectInfo [2 instances], getLayers, getCompositions) have empty or missing input schemas, automatically scoring them 0 on schema. Tool descriptions are present but mostly generic (10-40 chars), failing the 10-1024 character guideline and 20+ character floor. No parameter descriptions are provided for any tool. The source code is incomplete (uxp/pr/commands/index.js cuts off mid-function), making it impossible to verify actual tool registration or error handling. Tools are implemented across multiple disconnected plugin contexts (InDesign, Premiere Pro, After Effects) with duplicate 'getProjectInfo' names causing ambiguity.
Create a new InDesign document with specified margins, columns, page dimensions, and facing pages settings
Execute arbitrary ExtendScript code in After Effects and return the result as JSON
Get the active InDesign document's preferences including page dimensions, facing pages setting, margins, and column configuration
Get all compositions from the active After Effects project including id, name, dimensions, duration, and frame rate
Get all layers from the active After Effects composition including index, name, enabled state, selection state, and timing information
Get project information from After Effects including number of items, active item index, and project name
Get information about the active Premiere Pro project including name, path, id, and project content structure
Duplicate tool name 'getProjectInfo' across Premiere Pro and After Effects contexts causes irresolvable ambiguity. LLMs cannot distinguish which version to call and will guess, causing failures.
5 of 7 tools have NO input schemas (empty object '{}' with no properties defined). Tools affected: getActiveDocumentSettings, getProjectInfo (both), getLayers, getCompositions.
No parameter descriptions provided for ANY tool. Even createDocument and executeExtendScript, which have input schemas, lack descriptions of individual parameters (e.g., margins.bottom, margins.top, scriptString semantics).
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | F | 35 | 2026-07-28+ | v2 |
| 2026-03-09 | F | 25 | - | v1 |
No output schemas documented for any tool. LLMs cannot plan downstream calls or extract relevant fields. Only visible via code inspection, not in tool definitions.
executeExtendScript accepts arbitrary code execution with no documented timeout, resource limits, sandboxing, or permission gates. Violates security principle: gate destructive/sensitive tools behind permission checks.
Tools span three separate Adobe applications (InDesign, Premiere Pro, After Effects) but tool names do not disambiguate which app each belongs to. 'getProjectInfo' appears in two apps; 'getLayers' is AE-only but not prefixed. LLMs cannot route correctly.
Source code for uxp/pr/commands/index.js is truncated, preventing verification of actual tool registration, error handling, and return types for Premiere Pro tools.
No error handling or recovery guidance documented. Tools provide no indication of what errors are retryable, what the LLM should do if a tool fails, or how to disambiguate between recoverable and fatal failures.
createDocument and other write operations lack confirmation/dry-run patterns. An agent instructing the tool to create a document cannot preview the result before committing.
No pagination or result limiting documented for tools that return collections (getLayers, getCompositions). If results exceed token budget, LLMs have no guidance on how to constrain queries.