DOCSEARCHbyAlgolia
Submit Docs

wimpysworld Nix Configuration

repository·main·Indexed 20 days ago

https://github.com/wimpysworld/nix-config

A Nix-based configuration for managing Model Context Protocol (MCP) server configurations across AI agents such as Claude Code, Zed, and Codex. It implements a global coordination layer with a delegation-first model, utilizing a generated `delegate-task` skill for routing tasks to specialists. The system provides a type-safe way to distribute tool-use capabilities, defines a strict sub-agent response contract, and manages agent instructions and skills across various runtimes including OpenCode and Pi Agent.

Tokens
174.9K
Snippets
385
Records
773
Agent score
72%

What's inside wimpysworld-nix-config

  1. Overview of Communication Rules tripwire

    main

    The Communication Rules tripwire is a system designed to enforce brevity and clarity in AI agent outputs (Claude Code, Codex, Pi, and OpenCode). It uses a shared Python core to scan agent outputs for banned patterns (like em/en dashes or specific banned words) and applies enforcement policies based on the 'blast radius' of the output.

    Key Characteristics:

    • Deterministic Detection: Uses a stdlib-only Python scanner with exact matches rather than probabilistic LLM judging.
    • Two-Tier Enforcement: Differentiates between 'world-visible' output (which can be blocked) and 'prose' (which is corrected via silent re-issues).
    • Single Source of Truth: Nix generates a single rules fragment used by both the model (in prompts) and the gate (in the scanner).
    • Agent-Specific Wiring: Uses native mechanisms for each agent (e.g., command hooks for Claude Code/Codex, in-process plugins for Pi/OpenCode).

  2. Overview of AgentsView

    main

    AgentsView provides a shared dashboard for inspecting agent activity across multiple machines. It works by having local agents (running via Home Manager or Hermes) push their state into a central PostgreSQL database. A central server then serves a read-only web UI at /agentsview over the Tailnet Caddy host.

    Use AgentsView when you need to monitor agent activity across your fleet without logging into individual hosts or manually collecting local state files.

  3. Overview of available MCP Servers

    main

    The configuration provides several Model Context Protocol (MCP) servers across different transport methods (stdio, HTTP) and authentication types (login, bearer, OAuth).

    Unconditional Servers:

    • claude: Local binary via stdio. Used for Claude Code tools and agent delegation.
    • codex: Local binary via stdio. Used for Codex tools and agent delegation.
    • context7: Remote HTTP. Fetches live library documentation.
    • exa: Remote HTTP. Neural semantic web search and URL content extraction.
    • linear: Remote HTTP. Accesses Linear issues, projects, and comments.
    • slack: Remote HTTP (Conditional). Provides workspace Slack access.

    Note: claude and codex run as local binaries, while others use remote HTTP. slack is only available on hosts tagged workspace.

  4. Overview of Noughty

    main

    Noughty is a centralized system attributes module designed for NixOS, nix-darwin, and Home Manager flakes. It replaces the traditional method of threading computed values (like isLaptop, username, or hostname) through specialArgs with a typed, overridable NixOS-style options system under the noughty namespace.

    Key benefits include:

    • Type Safety: Prevents silent failures from typos in attribute names.
    • Overridability: Allows hosts to override central logic using standard NixOS mechanisms like lib.mkDefault and lib.mkForce.
    • Consistency: Provides a unified set of attributes across all module contexts (NixOS, Darwin, and Home Manager).
    • Centralized Logic: Eliminates triplicated logic for computing boolean flags across different builders.

  6. Overview of AI Agents and Specialists

    main

    The system utilizes a collection of specialized AI agents, each designed for a specific domain (e.g., security, testing, writing, implementation). Most agents inherit the model selected in the user's coding tool, with the exception of Garfield, which is a pinned agent using specific models (like sonnet for Claude Code or gpt-5.6-terra for others) to ensure deterministic, high-quality git workflow tasks.

    Key specialists include:

    • Rosey: Prompt & Skill Specialist (manages agent prompts and commands).
    • Batfink: Infrastructure Security Auditor.
    • Brain: Test Engineer.
    • Casper: Technical Writer (emulates a specific conversational British English voice).
    • Dibble: Code Security Auditor.
    • Donatello: Implementation Engineer (executes code changes).
    • Garfield: Git Workflow Specialist (enforces Conventional Commits).
    • Gonzales: Performance Specialist.
    • Penfold: Research Generalist (manages task lifecycles and research).
    • Penry: Code Reviewer (focuses on maintainability).
    • Velma: Documentation Architect.

  7. Overview of Hermes Cloudflare R2 backup scripts

    main

    The hermes-backup-r2 mixin provides three root-oriented helper commands for managing encrypted backups on Hermes-tagged hosts using Cloudflare R2. These commands handle the creation, verification, and restoration of encrypted backup archives and manifests.

    Available commands:

    • hermes-backup-r2: Creates and uploads an encrypted backup archive and manifest.
    • hermes-backup-verify-r2: Downloads and verifies a backup's metadata, hash, size, and structure.
    • hermes-restore-r2: Restores a specific backup to a designated destination directory.

  8. Use the love-system skill for system operations

    main

    The love-system skill provides access to information about the user's system. Use this skill when performing system operations, platform detection, system information retrieval, or managing system resources within LÖVE games.

    Common use cases:

    • Retrieving system and platform information.
    • Detecting operating system and hardware capabilities.
    • Accessing system-specific features.
    • Handling platform-specific behavior.
    • Managing system resources and capabilities.

  9. Use the love-audio skill for LÖVE audio operations

    main
    The love-audio skill provides an interface for audio-related operations within LÖVE games. Use this skill when you need to manage sound effects, music playback, audio recording, or immersive 3D audio experiences. It allows for managing multiple audio sources, setting listener properties (position, velocity, orientation), and applying audio effects.

  10. Use the love-keyboard skill for LÖVE game keyboard operations

    main
    The love-keyboard skill provides an interface to the user's keyboard within LÖVE games. Use this skill when you need to handle keyboard input, key events, text input, or implement keyboard-based game controls. It supports managing keyboard modifiers, special keys, and international keyboard layouts.