Agent Workflows

PDLC is the flagship Agentic workflow. Each stage has a clear agent role, artifact, and handoff to the next stage.

Discover

Goal: Understand the problem before writing code.

Agent does:
- Clarify users, goals, constraints, and success metrics
- Write docs/DISCOVER.md
- Avoid application code in this stage

Advance when: Discovery doc captures problem, users, and goals.

Design

Goal: Turn discovery into a buildable plan.

Agent does:
- Read docs/DISCOVER.md
- Produce architecture, UX flows, data model
- Write docs/DESIGN.md with an ordered implementation plan
- Uses planning prompts for step-by-step breakdown

Advance when: Design is actionable for development.

Develop

Goal: Implement against the design.

Agent does:
- Read design and discovery docs
- Create/modify application source files
- Run the project to validate progress
- Full filesystem and run tools available

Advance when: Core features match the design.

Test

Goal: Verify reliability.

Agent does:
- Run project/code and inspect output
- Diagnose and fix failures
- Write docs/TEST_REPORT.md
- Uses debugging prompts for root-cause analysis

Advance when: Checks pass or remaining risks are documented.

Deploy

Goal: Ship a preview.

Agent does:
- Confirm entry files and readiness
- Call run_project to publish preview
- Report preview URL (/deployed/{slug}/)

Advance when: Preview is live.

Iterate

Goal: Plan the next cycle.

Agent does:
- Capture feedback and improvement ideas
- Write docs/ITERATE.md with prioritized backlog
- Recommend returning to Discover for the next cycle

Advancing stages

In workspace AI Chat:
- Click a stage chip to switch
- Click Advance stage to move to the next stage with context
- Artifacts written in a stage mark it complete automatically

Other domains

These are registered stubs showing how the same engine expands beyond PDLC.