TasqLeveler
The TasqLeveler agent acts as an “Instruction Enhancer.” Its primary role is to read your project spec (tasq.md), analyze what you’re building, and automatically add the architectural guardrails that turn a 65% success rate into 95%+.
How it works
Before your first cycle even kicks off, TasqLeveler enhances your tasq.md with:
| Enhancement | What It Does | Why It Matters |
|---|---|---|
| 📦 Dependency Graph | Shows what can import what | No more circular import nightmares |
| 🎯 Golden Path Tests | Code that MUST work for each module | Success criteria, not vibes |
| 🧪 Mock Infrastructure | Mock servers for external services | Test without real APIs |
| 📋 Success Criteria | Global definition of “done” | Clear pass/fail, no ambiguity |
| ⏱️ Priority Guide | What to build if tokens run low | Smart budget allocation |
| 🔗 Base Classes | Abstract bases for similar modules | Consistent architecture |
The Numbers Don’t Lie
| Metric | Without TasqLeveler | With TasqLeveler |
|---|---|---|
| Imports resolve | 85% | 95%+ |
| Classes instantiate | 80% | 95%+ |
| Tests pass | 60% | 80%+ |
| Fully functional | 65% | 75-80% |
That’s a +15-20% improvement just from better instructions. The AI doesn’t get smarter - it gets better directions.
What’s configurable
config.yaml (under agents.tasqleveler):
provider(string): AI provider (e.g., ‘openai’, ‘anthropic’, ‘gemini’, ‘deepseek’, ‘qwen’)model(string): AI model (e.g., ‘gpt-4.1-mini’)