Agent♥︎Age
Catalog

NeuroDock Guardrail

Official

by tlennon-ie · TypeScript

Advisory detectors for rumination, hyperfocus, and sycophancy. Never silently blocks.

io.github.tlennon-ie/neurodock-mcp-guardrail MCP Server

The io.github.tlennon-ie/neurodock-mcp-guardrail Model Context Protocol (MCP) server provides clinical guardrails for NeuroDock. It detects three interaction patterns—rumination loops (OCD), unregulated hyperfocus (ADHD), and sycophancy/over-validation—using dedicated detectors that do not silently block.

🛠️ Key Features

  • Detects rumination, hyperfocus, and sycophancy/over-validation
  • Explicitly “Never silently blocks”
  • Detectors:
    • check_rumination — word-overlap Jaccard heuristic
    • check_hyperfocus — elapsed-threshold-with-end-of-day heuristic
    • check_sycophancy — four-pa* detector (truncated in provided excerpt)
  • AGPL topic listed among server topics

🚀 Use Cases

  • Advisory monitoring for neurodivergent interaction patterns (e.g., OCD/ADHD-related loops)
  • Integration with MCP-based NeuroDock workflows

⚡ Developer Benefits

  • Provides named detector functions (check_rumination, check_hyperfocus, check_sycophancy)
  • Uses heuristics documented in the README excerpt
  • References ETHICS.md and MANIFESTO.md for the full framework

⚠️ Limitations

  • README excerpt is truncated for check_sycophancy details
  • Full behavior beyond “advisory” and “never silently blocks” is not included in the provided data

Topics

accessibilityadhdagplanthropicaudhdautismbrowser-extensionclaudeclaude-desktopcognitive-substratedyslexiaexecutive-functionlocal-firstmcpmodel-context-protocolneurodivergentneurodiversityocdpythontypescript