Token-efficient tier-routed agent orchestrator — 40+ bundled skills, 19 subagent prompts, and Obsidian vault integration. Works with Claude Code, OpenCode, and Codex CLI.
Language
Self-contained. Zero external runtime dependencies. Drop into any project and go. MIT-licensed forever.
imlazy classifies every task into a tier — Quick, Standard, Heavy, Vault, or Learn — and loads only the workflow weight the task actually warrants. Simple edits get a fast path (~1.5K tokens). Complex features get the full pipeline. You save tokens on the majority of everyday work without losing rigor when it counts.
| Feature | What it means | |
|---|---|---|
| 🎯 | Tiered Routing | Tasks classified by complexity; each tier loads only relevant skills |
| 📦 | 40+ Workflow Skills | Planning, TDD, code review, brainstorming, debugging, API design, frontend/backend, security, WordPress + ACF, coding standards, and more |
| 🤖 | 19 Subagent Prompts | Planner, architect, code reviewer, security reviewer, TDD coach, build fixer, language reviewers (Python, Go, Rust, Java, C++, Kotlin, PHP) |
| 🏛️ | Obsidian Vault Integration | Markdown, Bases, Canvas, CLI, Defuddle skills + imlazy-vault for persistent knowledge across sessions |
| 🪶 | Token-Budgeted | Each skill stays within its budget; router loads only what the tier needs |
| 🔗 | Cross-Harness | Works with Claude Code, OpenCode, and Codex CLI |
| 🧩 | Self-Contained | No external plugin dependencies. Clone and run |
| Tier | When | Skills Loaded | Token Load |
|---|---|---|---|
| Quick | Single-file edit, typo, config bump, factual question | Inline execution + verify | ~1.5K |
| Standard | 1-3 file feature/bugfix, clear requirements | Planning (light) → TDD → Code Review | ~14.5K |
| Heavy | Multi-component, ambiguous scope, design request | Brainstorm → Planning (full, subagent dispatch) → TDD → Code Review | ~17K |
| Vault | Knowledge persistence | imlazy-vault (init, save, search, link) | vault skills |
| Learn | Router misclassification | imlazy-learn (log and adapt) | learn skill |
git clone https://github.com/hnikoloski/imlazy.git
cd imlazy
# Unix / WSL
bash install.sh
# Windows PowerShell
.\install.ps1bash tests/run.sh # Unix / WSL
python tests/run-tests.py # Any platform with Python 3imlazy activates via the /imlazy slash command or the i am lazy prefix:
/imlazy fix the typo on line 5 of README.md
i am lazy, add a /health endpoint that returns {status: ok}
i am lazy, vault init
The router classifies your request and loads only what the tier needs.
imlazy/
├── SKILL.md # Entry point / orchestrator
├── opencode.jsonc # OpenCode plugin config
├── .claude-plugin/ # Claude Code plugin manifest
├── .codex-plugin/ # Codex CLI plugin manifest
├── agents/ # 19 subagent prompt files
├── skills/ # 40+ SKILL.md workflow files
├── hooks/ # SessionStart hook
├── commands/ # /imlazy slash command
├── tests/ # Bash + Python test suite
├── mcp-configs/ # Bundled MCP server templates
├── docs/ # GitHub Pages homepage + SEO + llms.txt
└── assets/ # Logo, branding
| Skill | Purpose |
|---|---|
| imlazy-router | Task classification and tier dispatch |
| imlazy-bootstrap | Entry point, trigger detection |
| imlazy-planning | Implementation planning (light and full modes) |
| imlazy-tdd | Test-driven development with completion gate |
| imlazy-code-review | Code review request and receive workflow |
| imlazy-brainstorm | Requirements gathering and spec dialogue |
| imlazy-debug | Root-cause investigation workflow |
| imlazy-parallel | Cache-friendly parallel agent dispatch |
| imlazy-worktrees | Isolated workspace management |
| imlazy-learn | Router misclassification logging |
| imlazy-vault | Knowledge persistence and retrieval |
| imlazy-wordpress | WordPress + ACF theme/plugin/block development |
| imlazy-security | Security review patterns and vulnerability detection |
| imlazy-api-design | API design patterns and best practices |
| imlazy-frontend | Frontend development (React, TypeScript, etc.) |
| imlazy-backend | Backend development patterns |
| imlazy-standards | Coding standards reference |
| imlazy-verify | Pre-completion verification workflow |
| imlazy-subagents | Subagent-driven development workflow |
| imlazy-write-plans | Writing implementation plans |
| imlazy-write-skills | Writing and verifying skills |
| imlazy-receive-review | Receiving code review feedback |
| imlazy-request-review | Requesting code reviews |
| imlazy-strategic-save | Strategic context compaction |
| imlazy-eval | Evaluation-driven development |
| imlazy-finish-branch | Branch completion and merge workflow |
| obsidian-markdown | Obsidian Flavored Markdown |
| obsidian-bases | Obsidian Bases (.base) |
| json-canvas | JSON Canvas for visual thinking |
| obsidian-cli | Obsidian vault CLI interaction |
| defuddle | Clean markdown extraction |
19 subagent prompt files in agents/, dispatched by Heavy-tier planning:
Core: imlazy-planner, imlazy-architect, imlazy-code-reviewer, imlazy-security-reviewer, imlazy-tdd, imlazy-build-fixer, imlazy-e2e, imlazy-debugger, imlazy-docs, imlazy-refactor, imlazy-database, imlazy-docs-lookup
Language Reviewers: imlazy-python-review, imlazy-go-review, imlazy-rust-review, imlazy-java-review, imlazy-cpp-review, imlazy-kotlin-review, imlazy-php-review
| Harness | Config File | Install |
|---|---|---|
| Claude Code | .claude-plugin/plugin.json |
install.sh (auto-detect or --target claude) |
| OpenCode | opencode.jsonc |
Already in-repo — run opencode load |
| Codex CLI | .codex-plugin/codex.md |
install.sh --target codex |
# Unix / WSL (bash required)
bash tests/run.sh
# Any platform with Python 3
python tests/run-tests.pyTests verify: self-containment (no cross-plugin dependencies), router accuracy against fixtures, token budgets for every skill, and hook injection correctness.
MIT — see LICENSE and ATTRIBUTIONS.md for details.