This server exposes only 2 tools with severe definition quality gaps. Both tools accept arbitrary code execution inputs (command, script) with minimal validation, constraints, or error guidance. Descriptions are present but generic and do not convey the destructive nature of operations clearly enough for safe agent use. No parameter descriptions exist. No output schemas are documented. No pagination, structured response format, or error classification. The naming is acceptable (send_command, run_script use action verbs), but the tools themselves represent a critical security risk, arbitrary Python execution in FreeCAD context should never be exposed to untrusted agents without sandboxing, permission gates, and comprehensive audit trails.
Run an arbitrary Python script in FreeCAD context. Args: script: Python script to execute in FreeCAD Returns: JSON string containing the execution result
Send a command to FreeCAD and get document context information. Args: command: Command to execute in FreeCAD Returns: JSON string containing: - Command execution result - Current document information - Active objects and their properties - View state
Both tools accept arbitrary, unsanitized string input (command, script) with no constraints, enums, patterns, or length limits. LLMs can pass malicious payloads (path traversal, code injection, destructive FreeCAD operations) without guardrails.
No input validation, sanitization, or permission checks visible in source code (addon.py, freecad_bridge.py not provided). Arbitrary Python execution in FreeCAD context is a severe security risk. No audit trail mechanism documented.
Output schemas not documented. Description claims JSON with multiple fields (execution result, document info, objects, view state) but no schema provided. LLMs cannot reliably extract or chain results.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | F | 40 | 2026-07-28+ | v2 |
| 2026-03-09 | F | 20 | - | v1 |
Descriptions do not declare operations as destructive/state-modifying. Per pattern:command-tool, agents must know which calls can be retried and which have irreversible consequences. Both tools modify FreeCAD state but lack this critical metadata.
No error handling guidance. Descriptions do not explain what happens on failure, what errors are retryable, or what the LLM should do next. No recovery guidance.
No confirmation/dry-run mechanism for destructive operations. Per pattern:confirmation-request, irreversible operations like arbitrary script execution should support a dry-run or multi-turn confirmation before commit.
No input parameter descriptions. Neither tool documents what the 'command' or 'script' parameters should contain, valid ranges, or constraints. Baseline: 100% of A+ tools have parameter descriptions.
Tool descriptions are under 100 characters (send_command 97, run_script 54), risking incompleteness. Baseline average is 194 chars (p10=34, p90=392). These are at the floor and lack actionable detail.