S3YED Doing Tasks - Execution Core
The Rule
If a task exists, you MUST use the appropriate skill before acting.
User message received
↓
Might any skill apply? → YES (even 1% chance) → Invoke skill
↓ ↓
NO ← definitely not Announce: "Using [skill]"
↓ ↓
Respond Follow skill exactly
Skill Priority
When multiple skills could apply:
- Process skills first (brainstorming, debugging) - determine HOW
- Planning skills second (write-plan) - create roadmap
- Execution skills third (doing-tasks, dispatch-multiple-agents) - do the work
- Verification skills last (verify-task) - confirm completion
The Workflow
Standard Project Flow:
brainstorming → write-plan → doing-tasks → verify-task
↑ ↓
└────── refinement ←───────────┘
With Parallel Execution:
brainstorming → write-plan → dispatch-multiple-agents → verify-task
↓
doing-tasks (per subagent)
Red Flags - STOP and Check Skills
| Thought | Reality |
|---|---|
| "This is just a simple question" | Questions are tasks. Check for skills. |
| "I need more context first" | Skill check comes BEFORE clarifying. |
| "Let me explore first" | Skills tell you HOW to explore. |
| "I can check files quickly" | Check for skills first. |
| "I remember this skill" | Skills evolve. Read current version. |
| "This doesn't need a formal skill" | If a skill exists, use it. |
| "I'll just do this one thing first" | Check BEFORE doing anything. |
Execution Discipline
Do:
- Check for skills BEFORE any action
- Follow the plan exactly
- Mark tasks complete as you finish
- Report blockers immediately
- Ask for clarification when unclear
Don't:
- Skip skills because "it's simple"
- Add unplanned scope without approval
- Work silently for long periods
- Guess when unclear
- Let blockers sit unreported
User Communication Preferences (Seyed / Weblyfe)
When reporting results or answering questions:
- No unprompted option enumeration. Never offer "I can do A, B, or C" unprompted. Give the direct answer or take the action. If the user wanted options, they will ask.
- Concise bullet output. Default to bullet lists, not tables, detailed explanations, or verbose descriptions. Deliver the result, not the process.
- No "I would" / "I could" language. Take action directly or state the result. Do not describe what you would hypothetically do.
- No preambles or postambles. Skip "Great question!", "Let me help you with that", "To summarize" and similar filler. Start with the answer.
Integration with Clawlist
For long-running or infinite tasks:
doing-tasks → update ongoing-tasks.md → schedule next run
The doing-tasks skill executes, then updates the task tracking file for heartbeat monitoring.
Sub-Skills Reference
- brainstorming - Before any creative work
- write-plan - After design, before execution
- dispatch-multiple-agents - For parallel independent tasks
- verify-task - After completion
Example
User: "Build me a todo app"
Correct flow:
- "Using brainstorming skill to clarify requirements"
- Brainstorm: Ask questions, explore approaches
- "Using write-plan skill to create implementation plan"
- Write-plan: Create checkpoints and tasks
- "Using doing-tasks skill to execute"
- Execute: Work through plan
- "Using verify-task skill to confirm completion"
- Verify: Check against plan, get user approval
Incorrect:
- Jumping straight to coding without brainstorming
- Starting work without a plan
- Skipping verification