Create an execution plan from a requirement, spec, or task. Use when starting new work that needs structured breakdown into tasks.
Install
npx skillscat add weegigs/claude-kitbash/kick-off Install via the SkillsCat registry.
Workflow Kick-off
Create a detailed execution plan. Does NOT implement — use /execute after approval.
Prerequisites
Plan Mode
Kick-off requires plan mode. If not active, use the EnterPlanMode tool to switch.
Task Tracking
Detect availability:
command -v bd &> /dev/null && echo "beads" || echo "todowrite"- beads available: Use
bdcommands for full task management - beads unavailable: Use TodoWrite with naming convention (see references)
Knowledge Base
Detect ByteRover:
command -v brv &> /dev/null && [ -d ".brv" ] && echo "byterover" || echo "none"- byterover available: Query prior knowledge before planning
- byterover unavailable: Skip knowledge queries
Guardrails
- Favor straightforward implementations; add complexity only when required
- Keep plans tightly scoped to requested outcome
- Ask clarifying questions before creating tasks
- All subtasks must be completable in a single session
- Do NOT implement during kick-off
Process
Phase 1: Requirement Analysis
Parse input — identify source type:
- Free-text description → analyze directly
- Document reference → read and extract
- Beads task →
bd show <id> - Spec file → read from
.agent-os/specs/
Query prior knowledge (if byterover available):
brv query "What do we know about <topic>?" brv query "Any past issues with <area>?"Prior knowledge may reveal lessons learned, architectural decisions, or gotchas. Skip if byterover unavailable.
Explore context — understand:
- Related existing code and patterns
- Dependencies and constraints
- Similar prior implementations
Inject standards (if agent-os configured):
If .agent-os/standards/ exists: Invoke /standards-inject (auto-suggest mode)Ask clarifying questions (3-5 non-obvious):
- Edge cases, error handling, assumptions
- Performance, security, backwards compatibility
- Acceptance criteria and verification approach
Phase 2: Task Planning
Create parent task:
With beads:
bd create --title="<requirement summary>" --type=feature|task|bug --priority=2Without beads:
Add to TodoWrite: "[EPIC-XXX] <requirement summary>"Identify logical breakpoints:
- Group into phases (data layer, business logic, UI, etc.)
- Each phase produces a verifiable, working state
- Mark critical points requiring code review
Break into atomic subtasks:
- Each subtask independently verifiable
- Include acceptance criteria in description
- Add dependencies (beads:
bd dep add, TodoWrite: use ordering) - Insert checkpoint tasks at phase boundaries
- Add final task: "Final checkpoint: Full workflow verification"
Generate execution plan:
- Ordered subtask list with checkpoints
- Files likely modified per phase
- Testing approach
- Risks or blockers
Example structure:
Phase 1: Data Layer
- [EPIC/1] Add migration
- [EPIC/2] Implement repository
- [EPIC/✓] Checkpoint: Verify data layer
Phase 2: Business Logic
- [EPIC/3] Add validation service
- [EPIC/4] Implement commands
- [EPIC/✓] Checkpoint: Verify business logic
Phase 3: Integration
- [EPIC/5] Wire up components
- [EPIC/✓] Final checkpointPhase 3: Handoff
Present plan:
- Summary of requirement as understood
- Task IDs (beads) or TodoWrite items
- Execution order
- Remaining uncertainties
Save to TodoWrite:
- Full plan for agent continuity
- Enables
/executeto find and resume
On approval, remind user:
- Run
/executeto begin - Or start manually with first task
- Run
Completion Checklist
- Parsed input (free-text / document / beads / spec)
- Queried prior knowledge (if byterover available)
- Explored codebase context
- Injected relevant standards (if agent-os configured)
- Asked 3-5 clarifying questions
- Created parent task
- Identified phases and breakpoints
- Created subtasks with acceptance criteria
- Added dependencies/ordering
- Inserted checkpoint tasks
- Plan summary prepared
- Saved to TodoWrite
See Also
/execute— Execute the plan after approval/next— Find next task when resuming/check— Verification at checkpoints/standards-inject— Inject relevant coding standards (agent-os)