homeofe/improvements
Portable, LLM-agnostic framework for multi-model AI workflows: model routing, structured agent handoff (AAHP), and a phased agent pipeline. MIT.
ARCHETYPE
Orchestrator
A bench of specialised subagents. The main agent mostly delegates.
Copy this rig
# review before running: this installs third-party code
$ npx degit homeofe/improvements/.claude ./rig-improvements # inspect, then merge into .claude/ MCP servers are added to Claude Code at local scope; env vars are shown as YOUR_… placeholders — we never store values. Files are fetched with degit into a separate folder so you can review before merging.
$ curl -fsSL --create-dirs -o .claude/agents/architect.md https://raw.githubusercontent.com/homeofe/improvements/main/.claude/agents/architect.md $ curl -fsSL --create-dirs -o .claude/agents/auditor.md https://raw.githubusercontent.com/homeofe/improvements/main/.claude/agents/auditor.md $ curl -fsSL --create-dirs -o .claude/agents/handoff-manager.md https://raw.githubusercontent.com/homeofe/improvements/main/.claude/agents/handoff-manager.md $ curl -fsSL --create-dirs -o .claude/agents/implementer.md https://raw.githubusercontent.com/homeofe/improvements/main/.claude/agents/implementer.md $ curl -fsSL --create-dirs -o .claude/agents/researcher.md https://raw.githubusercontent.com/homeofe/improvements/main/.claude/agents/researcher.md $ curl -fsSL --create-dirs -o .claude/agents/reviewer.md https://raw.githubusercontent.com/homeofe/improvements/main/.claude/agents/reviewer.md
Merge into .claude/settings.json (project) or ~/.claude/settings.json (user). Hook commands reference scripts in the source repo — copy those too.
{
"permissions": {
"deny": [
"Bash(rm -rf*)",
"Bash(git push --force*)",
"Bash(git reset --hard*)"
]
}
} Subagents (6)
| architect model: opus | Use for system design decisions, architecture reviews, and ADR creation. AAHP Phase 2: Architecture Decision. |
| auditor model: opus | Use ON DEMAND (like /review-cycle) to audit grounding and trust-of-claims, and OPTIONALLY as Phase 4.5 (optional, pre-handoff) for high-impact tasks. Scope is provenance and grounding, NOT code review |
| handoff-manager model: haiku | Use at end of session to manage AAHP handoff state. Updates MANIFEST.json, validates checksums, manages task lifecycle. |
| implementer model: sonnet | Use for code implementation, writing tests, and executing ADR instructions. AAHP Phase 3: Implementation. |
| researcher model: sonnet | Use PROACTIVELY when task involves unknown libraries, APIs, or external services. AAHP Phase 1: Research & Context gathering. |
| reviewer model: opus | Use PROACTIVELY after code changes to auth, data handling, or security-sensitive areas. AAHP Phase 4: Review. |
Hooks (1)
| event | matcher | runs |
|---|---|---|
| PostToolUse | Edit|Write | bash scripts/validate-edit.sh |
Slash commands (6)
/challenge/handoff/next/review-cycle/route/status
Permissions
deny (3)
Bash(rm -rf*)
Bash(git push --force*)
Bash(git reset --hard*)
ask (0)
—
allow (11)
Read
Glob
Grep
Bash(git status*)
Bash(git diff*)
Bash(git log*)
Bash(sha256sum*)
Bash(wc -l*)
Bash(ls*)
Bash(cat*)
Bash(find*)
Similar rigs
OMOCHInoHOSHI/dotfiles-claude
.claude
Pragmatist 293 tok ·
Saba-Var/claude-ward
A read-only tripwire that watches Claude Code's local config and reports suspicious changes.
Pragmatist 1.8k tok ·
Trolzie/claude-loop-systems-demo
A small demo of Markdown-as-system AI dev loops with Claude Code: issues → agents → skills → verification.
Pragmatist 480 tok ·
nagamine-git/dotfiles
Chezmoi-managed dotfiles across 3 machines (Arch Linux — CachyOS/EndeavourOS — & macOS). Hyprland/Wayland, Zsh + Starship, Neovim, Ghostty, tmux, fcitx5, waybar
Pragmatist 1.4k tok ·