Domain: Foundations / Normalization
Status: Canonical
Task Normalization is the process of transforming ambiguous user intent into deterministic execution plans. In dev.kit, the AI agent acts as the primary Mapper, reasoning about the request and mapping it to the appropriate repository workflows and primitives.
The agent is responsible for dynamic prompt transformation. It receives intent from the user, identifies the required capabilities, and sends structured instructions to the dev.kit workflow engine.
Used for well-defined engineering tasks where the path is predictable and hardened.
- Example: Git Synchronization, environment hydration (
doctor), or diagram rendering. - Enforcement: Direct mapping to
lib/commands/ordocs/workflows/assets/*.yaml.
Used for creative or complex tasks where the agent must reason about the best path before committing to a sequence.
- Example: Implementing a new feature, refactoring complex logic, or resolving multi-domain drift.
- Enforcement: The agent generates a custom
workflow.mdthat orchestrates multiple primitives.
Agents are auto-mapped to send and receive context from repository workflows. If a task requires something outside of existing scripts or tools, the agent:
- Reasons about the implementation gap.
- Generates the necessary code or documentation patterns.
- Packages the resolution into a normalized
dev.kitworkflow step.
The normalization mapper routes common engineering intents to specialized UDX repositories:
| Intent Domain | Grounding Target | Mapping logic |
|---|---|---|
| Containerization | udx/worker |
Normalize to base environment specs. |
| Plugin Dev | udx/wp-stateless |
Normalize to structural plugin patterns. |
| CI/CD / Actions | udx/reusable-workflows |
Normalize to validated pipeline steps. |
Task normalization is operationalized through canonical UDX resources:
| Requirement | Grounding Resource | Role |
|---|---|---|
| Logic Mapping | udx/dev.kit |
Dynamic discovery and task resolution engine. |
| Context Hub | docs/workflows/README.md |
Source of truth for available repository sequences. |
| Fidelity | udx/worker |
Deterministic runtime for validating normalized plans. |
Normalization ensures high-fidelity execution through systematic pattern recognition:
- Observation-Driven Management: Optimizing task assignment through identified patterns.
- Autonomous Technical Operations: Principles for high-fidelity agent grounding and execution.
UDX DevSecOps Team