Back to skills
SKILL.md
2504 Design 23a5a3fd
ASecurity- **Feature**: enforce-workflow-boundaries - **Status**: APPROVED - **Created**: 2026-02-05 - **Author**: /zerg:design
- 9 stars
- 0 votes
- 0 copies
- 0 views
- Added October 11, 2026
Security analysis
100/100npx -y skills add tools-only/X-Skills --skill 2504-design_23a5a3fd --agent claude-codeAre you the author of 2504 Design 23a5a3fd?
Add the live security badge to your README. It updates with every re-scan.
[](https://www.skillsdirectory.com/skills/tools-only-2504-design-23a5a3fd)# Technical Design: enforce-workflow-boundaries
## Metadata
- **Feature**: enforce-workflow-boundaries
- **Status**: APPROVED
- **Created**: 2026-02-05
- **Author**: /zerg:design
---
## 1. Overview
### 1.1 Summary
Add explicit WORKFLOW BOUNDARY sections to `/z:plan` and `/z:brainstorm` commands preventing auto-progression to the next workflow phase. This is prompt-level enforcement via documentation changes only.
### 1.2 Goals
- Prevent `/z:plan` from ever auto-running `/z:design` or implementation
- Prevent `/z:brainstorm` from ever auto-running `/z:plan`
- Ensure user maintains explicit control over workflow progression
### 1.3 Non-Goals
- Programmatic enforcement in Python code
- Changes to `/z:design`, `/z:rush`, or other commands
- CI/CD validation of workflow boundaries
---
## 2. Architecture
### 2.1 High-Level Design
No architecture changes. This is documentation-only.
```
┌─────────────────┐ HARD STOP ┌─────────────────┐
│ /z:brainstorm │ ──────────────→ │ USER │
└─────────────────┘ │ (manual run) │
└────────┬────────┘
│
▼
┌─────────────────┐ HARD STOP ┌─────────────────┐
│ /z:plan │ ──────────────→ │ USER │
└─────────────────┘ │ (manual run) │
└────────┬────────┘
│
▼
┌─────────────────┐
│ /z:design │
└─────────────────┘
```
### 2.2 Component Breakdown
| Component | Responsibility | Files |
|-----------|---------------|-------|
| plan.md | WORKFLOW BOUNDARY section, simplified Phase 5.5 | zerg/data/commands/plan.md |
| plan.core.md | Same changes (keep in sync) | zerg/data/commands/plan.core.md |
| brainstorm.md | WORKFLOW BOUNDARY section, AskUserQuestion handoff | zerg/data/commands/brainstorm.md |
| brainstorm.core.md | Same changes (keep in sync) | zerg/data/commands/brainstorm.core.md |
### 2.3 Data Flow
N/A — Prompt documentation only.
---
## 3. Detailed Design
### 3.1 WORKFLOW BOUNDARY Section Template
Add at TOP of each command file (after title, before Flags):
```markdown
## ⛔ WORKFLOW BOUNDARY (NON-NEGOTIABLE)
This command MUST NEVER:
- Automatically run {next-command} or any {next-phase} phase
- Automatically proceed to {downstream-phase}
- Call the Skill tool to invoke another command
- Write code or make code changes
After Phase {N} completes, the command STOPS. The user must manually run {next-command}.
```
### 3.2 plan.md Changes
1. Add WORKFLOW BOUNDARY section after title (before ## Flags)
2. Update Phase 5.5 to remove "Run /z:design now" option (keep 2 options only)
3. Add ⛔ warning after Phase 5.5
### 3.3 brainstorm.md Changes
1. Add WORKFLOW BOUNDARY section after title (before ## Flags)
2. Update Phase 4 (Handoff) to use AskUserQuestion with 2 options
3. Add ⛔ warning after Phase 4
---
## 4. Key Decisions
### 4.1 Prompt-Level Enforcement Only
**Context**: Could enforce workflow boundaries in Python orchestrator code.
**Decision**: Prompt-level only
**Rationale**: The issue is Claude's interpretation, not code behavior. Direct instruction is the fix.
### 4.2 Remove "Run now" Option
**Context**: Phase 5.5 of plan.md has "Run /z:design now" which could confuse.
**Decision**: Remove it, keep only "Clear context, then /z:design" and "Stop here"
**Rationale**: Eliminates any ambiguity about auto-execution permission.
---
## 5. Implementation Plan
### 5.1 Phase Summary
| Phase | Tasks | Parallel | Est. Time |
|-------|-------|----------|-----------|
| Core | 4 | Yes (pairs) | 10 min |
### 5.2 File Ownership
| File | Task ID | Operation |
|------|---------|-----------|
| zerg/data/commands/plan.md | TASK-001 | modify |
| zerg/data/commands/plan.core.md | TASK-002 | modify |
| zerg/data/commands/brainstorm.md | TASK-003 | modify |
| zerg/data/commands/brainstorm.core.md | TASK-004 | modify |
### 5.3 Dependency Graph
```mermaid
graph TD
T001[TASK-001: plan.md]
T002[TASK-002: plan.core.md]
T003[TASK-003: brainstorm.md]
T004[TASK-004: brainstorm.core.md]
```
All tasks are independent (Level 1). No dependencies.
---
## 6. Risk Assessment
| Risk | Probability | Impact | Mitigation |
|------|-------------|--------|------------|
| Claude ignores HARD STOP | Low | Med | Use prominent ⛔ emoji and ALL CAPS |
| Files get out of sync | Low | Low | Tasks update both .md and .core.md |
---
## 7. Testing Strategy
### 7.1 Unit Tests
N/A — No code changes.
### 7.2 Integration Tests
N/A — No code changes.
### 7.3 Verification Commands
Manual verification:
- Read each modified file, confirm WORKFLOW BOUNDARY section present
- Confirm plan.md Phase 5.5 has only 2 options
- Confirm brainstorm.md Phase 4 has AskUserQuestion
---
## 8. Parallel Execution Notes
### 8.1 Safe Parallelization
- All 4 tasks are Level 1 (no dependencies)
- Each task owns exactly one file
- Can run all 4 in parallel
### 8.2 Recommended Workers
- Minimum: 1 worker
- Optimal: 2 workers (plan pair + brainstorm pair)
- Maximum: 4 workers
### 8.3 Estimated Duration
- Single worker: 10 min
- With 2 workers: 5 min
- Speedup: 2x
---
## 9. Approval
| Role | Name | Date | Signature |
|------|------|------|-----------|
| Architecture | | | PENDING |
| Engineering | | | PENDING |
Attribution
Comments
Loading comments…
Newsletter
Get the weekly list of safe new skills, and the ones we pulled.