---

name: oberplan description: Use when ANY planning, scoping, or design work is needed BEFORE implementation. Invoke when user asks to "build", "create", "implement", "add feature", "design", or describes desired functionality. CRITICAL - this skill dispatches FIRST, identifies relevant lens skills, clarifies requirements, and produces agent-executable plans. Triggers on "help me build", "I want to create", "let's implement", "design a", "plan for", "how should we approach", "figure out how to", or any request that requires understanding WHAT to build before building it.

Skill: oberplan

Meta-skill that orchestrates planning by loading domain-specific lens skills, clarifying requirements, and producing agent-executable plans.

The Iron Law

CopyNO PLAN SHIPS WITHOUT USER CONFIRMATION AND FINAL REVIEW

This applies to:

• "Quick" features
• "Obvious" implementations
• Plans that "should be straightforward"
• Your 5th plan after 4 successes

Skipping phases = agents execute wrong plans = rework.

---

Required Workflow

Copy1. Lens Selection (match user intent to domain skill)
      ↓
2. Requirements Clarification (guiding questions, expected outputs)
      ↓
3. Plan Construction (phase-by-phase OR finalized, per user preference)
      ↓
4. User Confirmation (explicit approval)
      ↓
5. Final Review (catch gaps before dispatch)
      ↓
6. Plan Quality Review (code review the plan itself)
      ↓
7. Save Plan to File (persist for execution)
      ↓
8. Output (agent-executable planning document)
      ↓
9. Generate Execution Prompt (oberprompt bootstrap for fresh context)
      ↓
10. Execution Handoff (provide bootstrap prompt for fresh context)

---

Phase 1: Lens Selection

Before ANY planning, identify the domain skill that provides the right "lens" for understanding this problem.

Process

1. Parse user request for domain signals
2. Match to available lens skills
3. Load the skill to inform subsequent phases

Domain-to-Skill Mapping

Output

CopyLENS SELECTED: [skill-name]
RATIONALE: [why this lens fits]
LOADING: [invoke skill to inform subsequent phases]

If no lens matches, proceed directly to Phase 2 with generic planning approach.

---

Phase 2: Requirements Clarification

Do NOT plan until requirements are understood.

Guiding Questions (ask in batches of 2-3)

Decision: Interactive vs Finalized

Output

CopyREQUIREMENTS CONFIRMED:
- Outcome: [what success looks like]
- Scope: [in/out boundaries]
- Constraints: [technical, timeline, dependencies]
- Mode: [Interactive | Finalized]

---

Phase 3: Plan Construction

Plan Structure

Every plan MUST include:

Checkpoints

See checkpoint-templates.md for checkpoint types, capability proofs, placement rules, and templates.

Interactive Mode

For each phase:

1. Present phase details
2. Ask: "Does this align with your expectations? Any adjustments?"
3. Incorporate feedback
4. Proceed to next phase

Finalized Mode

1. Construct complete plan
2. Present entire plan
3. Ask for single approval

Phase Template

See phase-templates.md for phase and document templates.

---

Phase 4: User Confirmation

MANDATORY - Do not proceed without explicit approval.

Confirmation Request

CopyPLAN COMPLETE

[Summary of phases]

Ready to proceed? Please confirm:
- [ ] Scope is correct
- [ ] Phase order makes sense
- [ ] Agent assignments are appropriate
- [ ] Nothing critical is missing

Reply "approved" to proceed to final review, or specify changes.

---

Phase 5: Final Review

After user approval, systematic gap check.

Review Checklist (Complete EVERY item)

Gap Detection

For each phase, ask:

• "What could go wrong here?"
• "What am I assuming that might not be true?"
• "What would an agent need to know that isn't explicitly stated?"

Output

CopyFINAL REVIEW COMPLETE:
- Gaps identified: [list or "none"]
- Mitigations added: [list or "n/a"]
- Plan status: READY FOR QUALITY REVIEW

---

Phase 6: Plan Quality Review

Code review the plan itself before execution.

Plans are code for agents. They have edge cases, implicit assumptions, and phases that may be too vague.

Review Process

Dispatch a review agent (via oberagent) to evaluate the plan:

CopyTask(
  subagent_type="general-purpose",
  description="Review: implementation plan quality",
  prompt="Review this implementation plan for completeness and executability.

  PLAN:
  [Full plan content]

  REVIEW CRITERIA:
  1. Phase objectives are actionable (not vague)
  2. Agent prompts have enough context to succeed
  3. Checkpoints have testable pass criteria
  4. Dependencies are explicit (no hidden assumptions)
  5. Risk mitigations are realistic
  6. Nothing requires information not available at that phase

  RETURN FORMAT:
  VERDICT: [READY | NEEDS_REVISION]

  If NEEDS_REVISION:
  ISSUES:
  - [Phase N] - [specific issue]

  SUMMARY: [1-2 sentences]"
)

---

Phase 7: Save Plan to File

Persist the plan for execution in this or a fresh context.

File Location

Copy~/.local/state/oberplan/plans/{project-name}-{timestamp}.md

File Format

Include: Plan content, execution instructions, and bootstrap prompt.

---

Phase 8: Output

Produce agent-executable planning document.

See phase-templates.md for the document format.

---

Phases 9-10: Bootstrap & Handoff

See bootstrap-guide.md for execution prompt generation and handoff procedures.

---

Red Flags - STOP and Reconsider

---

Integration

With oberexec

After plan approval, provide bootstrap prompt for fresh context execution. oberexec handles subagent-driven execution with checkpoints.

With oberagent

• Phase 6: Dispatch plan review agent via oberagent
• Execution: oberexec uses oberagent for all dispatches (in fresh context)

With oberprompt

• Phase 9: Use oberprompt principles to craft the bootstrap prompt for fresh context execution
• Ensures bootstrap prompt is outcome-focused, minimal constraints, high confidence

With oberdebug

If user request involves fixing bugs, redirect to oberdebug - debugging is not planning.

With code-foundations

• Phase 6: Plan review agent invokes code-foundations for quality evaluation
• Phases are designed with code-foundations principles (clear objectives, testable outputs)