Domain: AI / Workflows
Status: Canonical
Engineering Loops are the standardized execution plans used by agents to resolve Drift. By following these deterministic sequences, dev.kit ensures that complex tasks—from feature implementation to documentation synchronization—remain grounded in repository truth.
Every high-fidelity task follows the Analyze -> Normalize -> Process -> Validate -> Capture cycle.
Standard loop for implementing new capabilities with TDD and documentation.
- Goal: Expand repository "Skills" while maintaining 12-factor compliance.
- Steps:
- Analyze: Audit existing code and docs to identify the implementation gap.
- Normalize:
dev.kit task startto create a boundedworkflow.md. - Process:
dev.kit skills runto implement logic and test cases. - Validate:
dev.kit doctorto verify environment health and TDD success. - Capture:
dev.kit sync runto logically group and commit the resolution.
Deterministic lifecycle for identifying, reproducing, and resolving repository defects.
- Goal: Restore repository integrity with verified test evidence.
- Steps:
- Analyze:
dev.kit doctorto detect environment or software drift. - Normalize: Define reproduction steps in a new
workflow.md. - Process: Apply the fix and implement a regression test.
- Validate: Execute the test suite within the Worker Ecosystem.
- Capture:
dev.kit sync runto finalize the fix and update the Skill Mesh.
- Analyze:
Workflow for synchronizing repository documentation and agent context.
- Goal: Eliminate documentation drift and hydrate the Skill Mesh.
- Steps:
- Analyze: Scan
docs/and script headers for outdated metadata. - Normalize: Map documentation updates to current repository reality.
- Process:
dev.kit visualizerto regenerate high-fidelity architecture diagrams. - Validate: Verify that all internal and external links are high-fidelity.
- Capture:
dev.kit ai syncto ground the agent in the updated knowledge.
- Analyze: Scan
The standard engineering loops are operationalized through specialized UDX targets:
| Loop Domain | Grounding Target | Pattern Role |
|---|---|---|
| Logic Implementation | udx/dev.kit |
Intent normalization and task management. |
| Environment Parity | udx/worker |
Deterministic runtime for loop execution. |
| Automation Flow | udx/reusable-workflows |
Pattern baseline for sequence steps. |
Engineering loops are operationalized through deterministic UDX engines:
| Loop Type | Grounding Resource | Role |
|---|---|---|
| Engineering | udx/dev.kit |
Dynamic normalization and task management. |
| Automation | udx/reusable-workflows |
Pattern baseline for implementation steps. |
- Resume First: Before starting a new loop, agents must check for active tasks (
dev.kit task active). - Hygiene: Aborted or stagnant loops must be pruned (
dev.kit task cleanup) to prevent context noise. - Feedback: Every iteration must emit high-signal progress to the
feedback.mdartifact.
Standardized loops ensure predictable delivery and high-fidelity results:
- Little's Law for Flow: Managing cycle time and throughput through systematic sequences.
- Observation-Driven Management: Optimizing task assignment and execution through identified patterns.
UDX DevSecOps Team