USE when ready to execute a plan created by create-plan. Orchestrates parallel agent execution following the plan's dependency graph. Reads from .cattoolkit/planning/{slug}/ and updates progress in-place.
/plugin marketplace add Git-Fg/thecattoolkit/plugin install sys-builder@cattoolkitThis skill is limited to using the following tools:
Execute project plans with optimized parallel agent orchestration. This skill is the primary entry point for all plan execution workflows.
MANDATORY: Plans MUST be read from:
.cattoolkit/planning/{project-slug}/
Always ask user via AskUserQuestion to determine execution mode:
Question: How would you like to execute this plan?
Options:
1. Start from Phase 1 (fresh execution)
2. Resume from Phase {X} (continue where left off)
3. Execute all remaining phases
4. Refine/update the plan first
Based on response:
For each task in the phase, classify:
Sequential Tasks:
Parallel Tasks:
Background Tasks:
run_in_background: trueParallel Execution Pattern:
For parallel tasks 2, 3, 4:
- Send SINGLE message with 3 Task tool calls
- Each Task call spawns a worker agent
- Provide complete context to each agent including:
- BRIEF.md content
- DISCOVERY.md content
- Task-specific scope and action
- Verification criteria
Agent Prompt Template:
You are executing Task {N} of Phase {XX}.
## Context
{Include BRIEF.md summary}
{Include relevant DISCOVERY.md sections}
## Your Task
**Scope**: {files/directories}
**Action**: {what to do}
**Verify**: {how to verify}
**Done**: {acceptance criteria}
## Execution Rules
- Execute in UNINTERRUPTED FLOW mode
- Self-verify before completing
- Log any blockers to HANDOFF.md
- Do NOT ask user questions during execution
## Output
When complete, report:
- Files modified
- Verification results
- Any issues encountered
Verification Protocol:
For each completed task:
1. Run "Verify" command/check
2. Compare against "Done" criteria
3. If PASS: Mark task complete
4. If FAIL: Log failure, decide retry vs escalate
After each task:
After each phase:
On blocker:
Task 1 → Task 2 → Task 3
Execute one at a time, wait for completion.
[Task 1, Task 2, Task 3] (all at once)
Single message with multiple Task calls.
Task 1 → [Task 2, Task 3, Task 4]
Complete Task 1, then launch 2-4 simultaneously.
[Task 1, Task 2] + Task 3 (background)
Launch 1-2 foreground, 3 with run_in_background: true.
pending - Not startedin_progress - Currently executingcompleted - All tasks done and verifiedblocked - Waiting on external inputpending - Not startedin_progress - Agent executingcompleted - Verified donefailed - Verification failedblocked - External dependencyThis skill should be used when the user asks to "create a hookify rule", "write a hook rule", "configure hookify", "add a hookify rule", or needs guidance on hookify rule syntax and patterns.
Create distinctive, production-grade frontend interfaces with high design quality. Use this skill when the user asks to build web components, pages, or applications. Generates creative, polished code that avoids generic AI aesthetics.