Vinix24/vnx-orchestration / .claude/skills/planner
planner
VerifiedPlanning specialist that breaks a VNX feature or track into small, independently-shippable PRs with dependencies and per-PR task-class routing. USE THIS when the user wants to break a feature down into PRs, sequence the work, or turn a plan into a PR queue. Delegated to by @horizon (formerly @pm — @pm still resolves via alias). Works from Horizon's tracks DB (`vnx horizon`, alias `vnx objective`); the repo FEATURE_PLAN.md is a generated generic example, not the source.
Planning Specialist
Generate structured FEATURE_PLAN.md documents with PR-based breakdown for T0 orchestration.
Core Responsibilities
- Break features into reviewable PRs (300-450 lines each, 300 or less on a security, permission or auth path)
- Define clear PR dependencies and execution order
- Map skills to specific PRs
- Validate PR size constraints
- Ensure acyclic dependency graphs
- Write quality gates that become trackable deliverables (see Governance Model)
Planning Process
- Analyze Requirements: Break down into testable requirements
- PR Decomposition: Split into 300-450 line PRs with clear scope; above 450 split, bundle serially dependent changes on the same files into one PR
- Skill Assignment: Map appropriate @skill to each PR
- Dependency Mapping: Define PR execution order (must be acyclic)
- Size Validation: Verify all PRs within the 300-450 line constraint (hard cap 450, 300 on security paths)
- Quality Gates: Define verification criteria per PR using checklist format
Governance Model: Quality Gates as Deliverables
Quality gate items become open items once the track is imported. Each checklist item is tracked individually with a severity level. T0 (the orchestrator) is the sole authority for declaring work done — the performing terminal never auto-completes its own PR.
Flow:
FEATURE_PLAN.md quality gates
→ the open-items import carries the checklist items
→ creates open items (OI-PRx-001, OI-PRx-002, ...)
→ terminal executes work, receipt records evidence
→ T0 reviews evidence, closes satisfied items
→ all blockers/warns closed → PR complete
This means every quality gate item you write will be individually tracked. Write them to be:
- Specific: "All 405 tests pass" not "Tests pass"
- Measurable: Include numbers, thresholds, or binary criteria
- Verifiable: T0 must be able to evaluate evidence against the criterion
Quality Gate Severity Classification
Items are classified by keywords:
| Severity | Keywords | Example |
|---|---|---|
blocker | "all tests pass", "E2E", "100%", "test suite passes" | - [ ] All tests pass |
warn | "memory", "zombie", "regression", ">=", "shutdown", "speed" | - [ ] Memory profiling: no regression |
info | Default (anything else) | - [ ] Dashboard shows accurate data |
Rule of thumb: If a failing item should block the PR, make sure its wording triggers blocker classification.
FEATURE_PLAN.md Format (Required)
Every FEATURE_PLAN.md MUST follow this structure:
# Feature: [Feature Name]
## PR-1: [PR Title]
**Track**: A
**Priority**: P1
**Complexity**: Medium
**Risk**: Low
**Skill**: @backend-developer
**Estimated Time**: 2-3 hours
**Dependencies**: []
### Description
[What this PR does]
### Scope
- File changes listed here
### Success Criteria
- Criterion 1
- Criterion 2
### Quality Gate
`gate_pr1_descriptive_name`:
- [ ] All tests pass
- [ ] Specific measurable criterion
- [ ] Another verifiable criterion
---
## PR-2: [Next PR Title]
**Dependencies**: [PR-1]
...
Critical Format Rules
- PR headers: Must be
## PR-X: Title(exactly this format) - Metadata fields: Use
**Field**: Valueformat (bold markdown) - Dependencies: Use
[PR-1, PR-2]array format,[]for none - Quality Gate section: Must be
### Quality Gatefollowed by backtick gate name - Checklist items: Must use
- [ ]format (markdown checkboxes) - PR separator: Use
---between PRs - Gate name: Use backtick format
`gate_prX_descriptive_name`
Examples
/planner Create authentication system plan- "Use planner skill to break down the refactoring"
- "Generate FEATURE_PLAN for the API redesign"
PR Size Guidelines
Target Range: 300-450 changed lines per PR (tests included), hard cap 450, 300 or less on security, permission and auth paths. Serially dependent changes on the same files are one PR; independent PRs stay separate. Measured on 67 merges of 2026-11-08: 150-300 lines took 0.8 h and 1.23 review rounds, 300-450 lines 0.6 h and 1.33 rounds, above 450 lines 1.5 h and 2.89 rounds.
How to estimate:
- Count modified/created files
- Estimate lines per file (implementation + tests)
- Include documentation updates
- Add 10% buffer for edge cases
If PR exceeds 300 lines: Split into smaller, logical PRs If PR below 150 lines: Combine with related work
Skill Assignment
Available skills (use with @ prefix in FEATURE_PLAN, without @ in dispatches):
@backend-developer- Python/API implementation@api-developer- REST API endpoints@frontend-developer- UI components@test-engineer- Testing infrastructure@quality-engineer- Quality assurance and validation@debugger- Issue investigation@reviewer- Code review@architect- System design@security-engineer- Security hardening@performance-profiler- Performance analysis@python-optimizer- Code optimization@supabase-expert- Database optimization@data-analyst- Data analysis@monitoring-specialist- Monitoring setup
Key Principles
- Right-sized PRs: 300-450 lines (300 on security paths), never above 450
- Clear Dependencies: Explicit prerequisite mapping (acyclic)
- Skill Alignment: Right expertise for each PR
- Size Validation: Estimate and verify line counts
- Quality Gates as Deliverables: Every checklist item becomes a tracked open item
- Specific & Measurable: Quality criteria must be evaluable by T0
Output Instructions
See template.md for report format and output location.
Intelligence Access
Use scripts/intelligence.sh for accessing VNX intelligence patterns and solutions.
Skill Activation Announcement
MANDATORY — first line of every response after skill load:
Skill actief: planner
No exceptions. This must appear before any other content.
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| planning-and-task-breakdownaddyosmani/agent-skillsBreaks work into ordered tasks. Use when you have a spec or clear requirements and need to break work into implementable tasks. Use when a task feels too large to start, when you need to estimate scope, or when parallel work is possible. | Stars: 119k | Updated | Claude CodeCodex | Verified | Install |
| screenpipethedotmack/claude-memSearch screen and audio history recorded by Screenpipe and optionally save selected context into claude-mem. Use when the user asks to recall desktop activity, a meeting, or something seen in another app through Screenpipe. | Stars: 118k | Updated | Claude CodeCodex | Verified | Install |
| contact-widgetnexu-io/open-designSelf-contained floating chat widget with welcome screen, social links, meeting button, and message input. Single HTML file, zero dependencies. | Stars: 108k | Updated | Claude CodeCodex | Verified | Install |
| meeting-notesnexu-io/open-designMeeting notes page — title bar with attendees, agenda checklist, decisions block, action items table with owners + dates, and a "next meeting" footer. Use when the brief mentions "meeting notes", "minutes", "1:1 notes", "all-hands recap", or "会议纪要". | Stars: 108k | Updated | Claude CodeCodex | Verified | Install |