workflow-orchestration

name

workflow-orchestration

description

Plan-mode-first execution and workflow orchestration. Automatically enters PLAN mode for non-trivial tasks (3+ implementation steps, architectural decisions, infrastructure planning, scalability concerns, multiple modules, integrations, complex workflows, or UX systems). Generates detailed execution plans, risks, dependencies, verification strategies. Stops and re-plans on failure. Use when starting any non-trivial task, before touching code, or when work behaves unexpectedly.

Workflow Orchestration

The Default Execution Loop

For ANY non-trivial task:

Research → Think → Plan → Build → Test → Verify → Validate → Reflect → Improve → Repeat

Never stop at "it works." Iterate until the solution feels elegant, scalable, intuitive, consumer-grade, and genuinely delightful. Every iteration should improve UX, performance, architecture, clarity, reliability, or user satisfaction.

Step	Output
Research	Competitor scan, prior art, modern UX patterns, constraints
Think	Architecture options, tradeoffs, "is there a more elegant solution?"
Plan	tasks/todo.md with milestones, risks, verification strategy
Build	Atomic-per-logical-change commits, conventional commit messages
Test	Unit + integration + edge cases — written before or with the code
Verify	Run tests, lint, type-check, UX walkthrough, regression check
Validate	"Would real users at scale love this?" Accessibility, perf, security
Reflect	What worked, what didn't, what felt weak, what introduced risk
Improve	Refactor weak areas, redesign hacky paths, polish UX
Repeat	Next iteration — never accept first-pass quality as final

After every correction from the user, append a lesson to tasks/lessons.md (see task-management-loop skill). Never repeat the same mistake twice.

Plan Mode Default

For ANY non-trivial task, enter PLAN mode first (call SwitchMode with target_mode_id: "plan").

What Is Non-Trivial?

A task is non-trivial if it includes any of:

• 3+ implementation steps
• Architectural decisions
• Infrastructure planning
• Scalability concerns
• Multiple modules
• External integrations
• Complex workflows
• UX systems design

Plan Contents

Before implementation, generate:

1. Detailed execution plan — ordered steps with file/symbol targets
2. Risk identification — what can fail, what's unknown
3. Dependency mapping — internal/external blockers
4. Verification strategies — how each step will be proven
5. Architecture decisions with rationale — ADRs for non-trivial choices

Failure Protocol

If anything fails or behaves unexpectedly:

1. STOP immediately
2. Re-evaluate assumptions
3. Re-plan before continuing
4. Do NOT continue blindly

Plan Mode Applies To

Implementation, debugging, verification, optimization, refactoring, deployment, testing.

Always reduce ambiguity upfront.

Task File Convention

Maintain tasks/todo.md as the single source of truth for current work. See task-management-loop skill for the full format and lifecycle.

Output Deliverables For Significant Features

Before implementation of significant features, generate:

• PRD (Product Requirements Document)
• TRD (Technical Requirements Document)
• Architecture diagrams (Mermaid)
• UX flows
• System flows
• Database schemas
• API contracts
• Folder structures
• Component inventories
• Deployment architecture
• Sprint plans
• Technical risk analysis
• Verification strategies

Then implement iteratively, validate thoroughly, and continuously improve.

Anti-Patterns

• Jumping into code without a plan
• Continuing past a failure instead of re-planning
• Treating "I think I know what to do" as sufficient planning
• Skipping risk analysis on infrastructure or architecture changes