Add Grok Build agent config adapted from Claude Code setup

Wire shared AGENTS.md and Claude settings through home-manager, plus a
native Grok config (always-approve, rose-pine) and a `gb` yolo alias.

AGENTS.md

@@ -0,0 +1,15 @@
+# Project notes for agents
+
+Deliberate decisions in this repo - do NOT silently revert them:
+
+- `homebrew.onActivation.cleanup = "zap"` in `configuration.nix` is intentional. It forces the good habit of declaring every Homebrew package in the Nix config instead of installing things ad-hoc, which keeps the machine reproducible. Do not soften it to `uninstall` or `none`. Users are warned about its effect; this note is for anyone tempted to change the setting itself.
+- Agent configs under `home/` are the source of truth. `home.nix` installs them with `mkOutOfStoreSymlink`, so edits to files in `home/` apply live after the first rebuild that creates the symlink. Prefer editing those paths, not the live `~/.claude` / `~/.grok` copies.
+- Grok Build is first-class here: `home/.grok/config.toml` and `home/AGENTS.md` (linked as `~/.grok/AGENTS.md`). Claude, Codex, and opencode still share the same `home/AGENTS.md` for multi-agent work.
+- Shell aliases: `cc` (Claude skip-permissions), `co` (Codex full-auto), `gb` (Grok --yolo). They are intentional high-agency shortcuts.
+
+## Maintaining this file
+
+Keep this file for knowledge useful to almost every future agent session in this project.
+Do not repeat what the codebase already shows; point to the authoritative file or command instead.
+Prefer rewriting or pruning existing entries over appending new ones.
+When updating this file, preserve this bar for all agents and keep entries concise.

home.nix

@@ -34,8 +34,10 @@ in
       push = "git push";
       pull = "git pull";
       m = "git switch main";
+      # High-agency agent launchers (same idea across tools)
       cc = "claude --dangerously-skip-permissions";
       co = "codex --full-auto";
+      gb = "grok --yolo";
     };
   };
 
@@ -52,11 +54,29 @@ in
     };
   };
 
-  # Edit-in-place: the real file stays in my repo, ~/.config just points at it.
+  # Edit-in-place: the real file stays in my repo, the live path just points at it.
   home.file.".config/wezterm".source =
     config.lib.file.mkOutOfStoreSymlink "${dotfiles}/home/.config/wezterm";
   home.file.".config/nvim".source =
     config.lib.file.mkOutOfStoreSymlink "${dotfiles}/home/.config/nvim";
-  home.file.".config/herdr".source = 
+  home.file.".config/herdr".source =
     config.lib.file.mkOutOfStoreSymlink "${dotfiles}/home/.config/herdr";
+
+  # Claude Code settings (also read by Grok for permissions/compat)
+  home.file.".claude/settings.json".source =
+    config.lib.file.mkOutOfStoreSymlink "${dotfiles}/home/.claude/settings.json";
+
+  # Shared agent policy - one file, many harnesses
+  home.file.".claude/CLAUDE.md".source =
+    config.lib.file.mkOutOfStoreSymlink "${dotfiles}/home/AGENTS.md";
+  home.file.".codex/AGENTS.md".source =
+    config.lib.file.mkOutOfStoreSymlink "${dotfiles}/home/AGENTS.md";
+  home.file.".config/opencode/AGENTS.md".source =
+    config.lib.file.mkOutOfStoreSymlink "${dotfiles}/home/AGENTS.md";
+  home.file.".grok/AGENTS.md".source =
+    config.lib.file.mkOutOfStoreSymlink "${dotfiles}/home/AGENTS.md";
+
+  # Grok Build native config
+  home.file.".grok/config.toml".source =
+    config.lib.file.mkOutOfStoreSymlink "${dotfiles}/home/.grok/config.toml";
 }

home/.claude/settings.json

@@ -0,0 +1,7 @@
+{
+  "theme": "dark-ansi",
+  "statusLine": {
+    "type": "command",
+    "command": "input=$(cat); model=$(echo \"$input\" | jq -r '.model.display_name'); used=$(echo \"$input\" | jq -r '.context_window.used_percentage // empty'); if [ -n \"$used\" ]; then printf \"%s | ctx: %.0f%% used\" \"$model\" \"$used\"; else printf \"%s\" \"$model\"; fi"
+  }
+}

home/.grok/config.toml

@@ -0,0 +1,28 @@
+# Grok Build user config - managed via home-manager (edit here, then rebuild if needed).
+# Adapted from Claude Code setup in kunchenguid/dotfiles + local preferences.
+#
+# Mapping notes:
+#   Claude --dangerously-skip-permissions  ->  permission_mode = "always-approve" / --yolo
+#   Claude theme dark-ansi + rose-pine stack ->  theme = "rosepine"
+#   Claude statusLine (model | ctx %)      ->  no direct equivalent; TUI shows this natively
+#   Shared AGENTS.md                       ->  ~/.grok/AGENTS.md (via home.nix)
+
+[marketplace]
+official_marketplace_auto_installed = true
+
+[[marketplace.sources]]
+name = "xAI Official"
+git = "https://github.com/xai-org/plugin-marketplace.git"
+
+[models]
+default = "grok-build"
+
+[ui]
+theme = "rosepine"
+max_thoughts_width = 120
+fork_secondary_model = "grok-build"
+# High-agency default, same idea as `claude --dangerously-skip-permissions` / `codex --full-auto`.
+# Toggle per session with Ctrl+O or /always-approve. Shell alias `gb` also launches with --yolo.
+permission_mode = "always-approve"
+yolo = false
+compact_mode = false

home/AGENTS.md

@@ -0,0 +1,14 @@
+# global agent instructions
+
+- Never use the em dash "—". Use plain dash "-" instead
+- When writing commit messages, NEVER auto-add your agent name as co-author
+- Never manually modify CHANGELOG.md files or any files that are marked as auto-generated
+- When making technical decisions, do not give much weight to development cost.
+  Instead, prefer quality, simplicity, robustness, scalability, and long term maintainability.
+- When doing bug fixes, always start with reproducing the bug in an E2E setting as closely aligned with how an end user would experience it as possible.
+  This makes sure you find the real problem so your fix will actually solve it.
+- When end-to-end testing a product, be picky about the UI you see and be obsessed with pixel perfection.
+  If something clearly looks off, even if it is not directly related to what you are doing, try to get it fixed along the way.
+- Apply that same high standard to engineering excellence: lint, test failures, and test flakiness.
+  If you see one, even if it is not caused by what you are working on right now, still get it fixed.
+- Before using "dynamic workflows", "ultra code" or any harness feature that immediately spawns a large swarm of subagents, always explain the tradeoffs and ask the user for explicit approval.