Deep Interview

<Why_This_Exists> AI can build anything. The hard part is knowing what to build. OMA's autopilot Phase 0 expands ideas into specs via analyst + architect, but this single-pass approach struggles with genuinely vague inputs. It asks "what do you want?" instead of "what are you assuming?" Deep Interview applies Socratic methodology to iteratively expose assumptions and mathematically gate readiness, ensuring the AI has genuine clarity before spending execution cycles. </Why_This_Exists>

<Execution_Policy>

• Ask ONE question at a time -- never batch multiple questions
• Target the WEAKEST clarity dimension with each question
• Make weakest-dimension targeting explicit every round: name the weakest dimension, state its score/gap, and explain why the next question is aimed there
• Gather codebase facts via Explore agent BEFORE asking the user about them
• Score ambiguity after every answer -- display the score transparently
• Do not proceed to execution until ambiguity ≤ threshold (default 0.2)
• Allow early exit with a clear warning if ambiguity is still high
• Persist interview state for resume across session interruptions
• Challenge agents activate at specific round thresholds to shift perspective </Execution_Policy>

Phase 1: Initialize

1. Parse the user's idea from {{ARGUMENTS}}
2. Detect brownfield vs greenfield:
   • Run Explore agent (haiku): check if cwd has existing source code, package files, or git history
   • If source files exist AND the user's idea references modifying/extending something: brownfield
   • Otherwise: greenfield
3. For brownfield: Run Explore agent to map relevant codebase areas, store as codebase_context
4. Initialize state via state_write(mode="deep-interview"):

{
  "active": true,
  "current_phase": "deep-interview",
  "state": {
    "interview_id": "<uuid>",
    "type": "greenfield|brownfield",
    "initial_idea": "<user input>",
    "rounds": [],
    "current_ambiguity": 1.0,
    "threshold": 0.2,
    "codebase_context": null,
    "challenge_modes_used": [],
    "ontology_snapshots": []
  }
}

5. Announce the interview to the user

Phase 2: Interview Loop

Repeat until ambiguity ≤ threshold OR user exits early:

Step 2a: Generate Next Question

Build the question generation prompt with:

• The user's original idea
• All prior Q&A rounds (conversation history)
• Current clarity scores per dimension (which is weakest?)
• Challenge agent mode (if activated -- see Phase 3)
• Brownfield codebase context (if applicable)

Question targeting strategy:

• Identify the dimension with the LOWEST clarity score
• Generate a question that specifically improves that dimension
• State, in one sentence before the question, why this dimension is now the bottleneck to reducing ambiguity
• Questions should expose ASSUMPTIONS, not gather feature lists
• If the scope is still conceptually fuzzy (entities keep shifting, the user is naming symptoms, or the core noun is unstable), switch to an ontology-style question

Question styles by dimension:

Step 2b: Ask the Question

Use AskUserQuestion with the generated question.

Step 2c: Score Ambiguity

After receiving the user's answer, score clarity across all dimensions using opus model (temperature 0.1).

Calculate ambiguity:

Greenfield: ambiguity = 1 - (goal × 0.40 + constraints × 0.30 + criteria × 0.30) Brownfield: ambiguity = 1 - (goal × 0.35 + constraints × 0.25 + criteria × 0.25 + context × 0.15)

Step 2d: Report Progress

Show the user their progress after each round:

Round {n} complete.

| Dimension | Score | Weight | Weighted | Gap |
|-----------|-------|--------|----------|-----|
| Goal | {s} | {w} | {s*w} | {gap or "Clear"} |
| Constraints | {s} | {w} | {s*w} | {gap or "Clear"} |
| Success Criteria | {s} | {w} | {s*w} | {gap or "Clear"} |
| Context (brownfield) | {s} | {w} | {s*w} | {gap or "Clear"} |
| **Ambiguity** | | | **{score}%** | |

Step 2f: Check Soft Limits

• Round 3+: Allow early exit if user says "enough", "let's go", "build it"
• Round 10: Show soft warning
• Round 20: Hard cap

Phase 3: Challenge Agents

At specific round thresholds, shift the questioning perspective:

Round 4+: CONTRARIAN Mode

Ask "What if the opposite were true?" or "What if this constraint doesn't actually exist?"

Round 6+: SIMPLIFIER Mode

Ask "What's the simplest version that would still be valuable?" or "Which constraints are actually necessary vs. assumed?"

Round 8+: ONTOLOGIST Mode (if ambiguity > 0.3)

Ask "What IS this, really?" to find the essence

Challenge modes are used ONCE each, then return to normal Socratic questioning.

Phase 4: Crystallize Spec

When ambiguity ≤ threshold (or hard cap / early exit):

1. Generate the specification using opus model with the full interview transcript
2. Write to file: .oma/specs/deep-interview-{slug}.md

Phase 5: Execution Bridge

After the spec is written, present execution options via AskUserQuestion:

Question: "Your spec is ready (ambiguity: {score}%). How would you like to proceed?"

Options:

1. Ralplan → Team (Recommended) — consensus-refine then execute with N coordinated agents
2. Ralplan → Ralph — consensus-refine then execute with persistence loop
3. Team directly — N coordinated parallel agents with the spec as shared plan
4. Ralph directly — persistence loop until all acceptance criteria pass
5. Refine further — continue interviewing

IMPORTANT: Invoke the chosen execution mode via Skill(). Do NOT implement directly.

<Tool_Usage>

• Use AskUserQuestion for each interview question
• Use Explore agent (haiku) for brownfield codebase exploration
• Use opus model (temperature 0.1) for ambiguity scoring
• Use state_write / state_read for interview state persistence
• Use Write tool to save the final spec to .oma/specs/
• Use Skill() to bridge to execution modes </Tool_Usage>

<Escalation_And_Stop_Conditions>

• Hard cap at 20 rounds: proceed with whatever clarity exists
• Soft warning at 10 rounds: offer to continue or proceed
• Early exit (round 3+): allow with warning if ambiguity > threshold
• User says "stop": stop immediately, save state for resume
• Ambiguity stalls: activate Ontologist mode to reframe </Escalation_And_Stop_Conditions>

Good Usage

Greenfield idea with no existing codebase:

User: "I want a tool that helps me track my learning goals"
OMA: [Deep Interview activates]
OMA: "What does 'track' mean to you in this context — a spreadsheet, an app, a CLI?"
User: "A CLI tool, something I run daily"
OMA: [Round 2: Focus on constraint clarity — online/offline, cross-platform?]
OMA: [Ambiguity: 45% → 22% after 6 rounds]
OMA: [Spec written to .oma/specs/learning-goal-tracker.md]
OMA: "Spec ready (ambiguity: 22%). How would you like to proceed?"

Brownfield extension requiring context questions:

User: "I want to add a recommendation engine to our existing MCP server"
OMA: [Deep Interview — brownfield detected]
OMA: [Explores existing MCP server codebase via oma-explorer]
OMA: "I found your server structure in plugins/mcp/. Should the recommendation engine be a new plugin or integrated into the existing handler pipeline?"
User: "New plugin — I want it to be optional"
OMA: [Context clarity improves — ambiguity drops]

Bad Usage

Detailed spec provided — skip to execution:

User: "Use the PRD in .oma/specs/auth-overhaul.md to implement the auth refactor"
OMA: [Deep Interview activates unnecessarily]
OMA: "What does 'auth refactor' mean to you?"
User: "Just use the PRD — I already wrote it out"

→ Use oma-ralph or oma-team with the existing spec directly.

User wants a quick brainstorm:

User: "Give me some ideas for a new CLI tool"
OMA: [Deep Interview starts scoring dimensions and asking Socratic questions]
OMA: "Round 1: Goal clarity is 30% — let's start there. What is the core problem..."
User: "Whoa, I just wanted ideas, not a full spec"

→ Use oma-planner for exploration/brainstorming, not deep-interview. </Examples>

<Final_Checklist>

• Interview state written to .oma/state.json with mode deep-interview
• Brownfield vs. greenfield detection confirmed (brownfield: codebase context gathered via oma-explorer)
• Ambiguity score calculated and shown to user after each round (formula applied correctly)
• Questions target the weakest dimension explicitly (named and justified in output)
• Challenge modes activated at correct round thresholds (Contrarian at 4+, Simplifier at 6+, Ontologist at 8+ if needed)
• Soft warning shown at round 10, hard cap enforced at round 20
• Spec written to .oma/specs/deep-interview-{slug}.md with all resolved dimensions captured
• Execution bridge presented to user with all 5 options and Skill() invocation for chosen mode
• If early exit taken, ambiguity score and remaining gaps are explicitly documented in the spec header
• Output format matches the Round progress table (Dimension | Score | Weight | Weighted | Gap) </Final_Checklist>