Arness Brainstorming
Walk the user through the entire Arness greenfield exploration pipeline in a single continuous session. The wizard invokes each pipeline skill in sequence using the Skill tool, only pausing at genuine decision gates where user input is needed. Conversational skills (discover, arch-vision) have internal feedback loops and flow without wizard-level gates.
This skill is a sequencer and decision-gate handler. It MUST NOT duplicate sub-skill logic. All pipeline work is done by the invoked skills. The wizard handles: entry routing, transitions between skills, progress display, and resumability.
Step 0: Ensure Configuration
Read ${CLAUDE_PLUGIN_ROOT}/skills/arn-spark-ensure-config/references/step-0-fast-path.md and follow its instructions. This guarantees a user profile exists and ## Arness is configured with Arness Spark fields before proceeding.
After Step 0 completes, extract from ## Arness:
- Vision directory, Use cases directory, Prototypes directory, Spikes directory, Visual grounding directory, Reports directory
Decision Gates
The wizard pauses at exactly 9 gates. Everything else flows automatically.
| Gate | When | Question | Options |
|---|---|---|---|
| G1 | Entry (artifacts detected) | "It looks like you have an in-progress greenfield pipeline. Resume or start fresh?" | Resume from [detected stage] / Start fresh |
| G-Stress | After discover | "Run stress tests on the product concept?" then "Select which stress tests to run" | Yes/Skip; then multi-select: Interview, Competitive, Pre-Mortem, PR/FAQ |
| G-Name | After stress-test/concept-review | "Explore brand names for your product?" | Yes / Skip |
| G2 | After arch-vision | "Write use cases before scaffolding?" | Write use cases (recommended) / Write use cases with expert debate (teams) / Skip to scaffold |
| G3 | After scaffold | "Validate technical risks before visual design?" | Run spikes (recommended if risks found) / Skip to visual sketch |
| G4 | Pre-prototype | "How should prototypes be validated?" | Standard review cycles / Expert debate (teams) for both / Expert debate (teams) for clickable only |
| G5 | After clickable prototype | "Lock the prototype and extract features?" | Lock prototype then extract features (recommended) / Extract features without locking / Review prototype first |
| G6 | After feature extraction | "Transition to development pipeline?" | Yes, run /arn-planning / Not yet |
| G7 | Entry (add-feature intent + completed prototype) | "Adding a new feature. This will update use cases, iterate the prototype, and add to the feature backlog. Ready?" | Proceed / Choose which steps to include / Cancel |
Workflow
Step 0.5: Extract Configuration
- Read the project's CLAUDE.md
## Arnesssection (guaranteed to exist after Step 0). - Extract configuration fields needed for artifact detection:
- Vision directory — for detecting product concept, architecture vision, scaffold summary, spike results, style brief, visual direction, feature backlog
- Use cases directory — for detecting use case files
- Prototypes directory — for detecting static and clickable prototype reports
- Spikes directory — for detecting spike POC directories
- Visual grounding directory — for detecting captured design references
- Reports directory — for detecting stress test reports and concept review report
Step 1: Detect Pipeline State (Resumability)
Check which artifacts exist on disk to determine if the user is resuming a previous pipeline run. Check from most advanced to least advanced — first match wins:
| Artifact | Detected State | Resume Point |
|---|---|---|
User says "add feature" / "new feature" AND <prototypes-dir>/clickable/final-report.md exists | Add Feature mode | Start at Add Feature Mode (Step AF) |
<vision-dir>/features/feature-backlog.md | Feature extraction complete | Resume at G6 |
<prototypes-dir>/locked/LOCKED.md exists AND NO <vision-dir>/features/feature-backlog.md | Prototype locked, extraction pending | Resume at feature-extract directly (invoke arn-spark-feature-extract, skip G5 lock step — already locked) |
F-*.md files in <vision-dir>/features/ but NO <vision-dir>/features/feature-backlog.md | Feature extraction interrupted | Resume at feature-extract (invoke to complete/rebuild the Feature Tracker) |
<prototypes-dir>/clickable/final-report.md | Clickable prototype done | Resume at G5 |
<prototypes-dir>/static/final-report.md | Static prototype done | Resume at clickable prototype (Step 9b) |
<vision-dir>/style-brief.md | Style defined | Resume at static prototype (Step 9a) |
<vision-dir>/visual-direction.md | Visual direction chosen | Resume at style-explore (Step 8) |
<vision-dir>/spike-results.md | Spike complete | Resume at visual-sketch (Step 7) |
<vision-dir>/scaffold-summary.md | Scaffold built | Resume at G3 |
UC-*.md in <use-cases-dir>/ | Use cases written | Resume at scaffold (Step 5) |
<vision-dir>/architecture-vision.md | Architecture defined | Resume at G2 |
<reports-dir>/stress-tests/concept-review-report.md AND <vision-dir>/naming-brief.md with "Final Decision" section populated | Concept review + naming complete | Resume at arch-vision (Step 3) |
<reports-dir>/stress-tests/concept-review-report.md AND <vision-dir>/naming-brief.md exists but no "Final Decision" section | Naming in progress | Resume at G-Name (naming skill handles internal resume) |
<reports-dir>/stress-tests/concept-review-report.md AND NO <vision-dir>/naming-brief.md | Concept review complete, naming not started | Resume at G-Name |
Any stress test report (interview-report.md, competitive-report.md, premortem-report.md, prfaq-report.md) in <reports-dir>/stress-tests/ but NO concept-review-report.md | Stress tests run, review pending | Resume at concept review offer within G-Stress |
<vision-dir>/product-concept.md | Product discovered | Resume at G-Stress |
| Greenfield fields but no artifacts | Greenfield initialized | Begin at discover (Step 2) |
If artifacts are detected:
Show the detected state and ask:
Ask (using AskUserQuestion):
It looks like you have an in-progress greenfield pipeline at [detected stage]. Would you like to resume or start fresh?
- Resume from [detected stage] — Skip to the detected resume point
- Start fresh — Begin at Step 2 (existing artifacts are preserved)
If ambiguous or conflicting artifacts are detected (e.g., architecture-vision.md exists but no product-concept.md, or multiple conflicting states): list the detected artifacts and ask the user to clarify which state to resume from.
If no artifacts detected: skip G1 and proceed directly to Step 2.
Step 2: Product Discovery (Automated)
Show progress:
Spark Pipeline: DISCOVER --> stress-test --> naming --> arch-vision --> use-cases --> scaffold --> spike --> visual-sketch --> style-explore --> prototypes --> feature-extract
^^^^^^^^
Inform the user: "Let's start by shaping your product idea..."
Skill: arn-spark:arn-spark-discover
The discover skill has its own internal conversation loop with the product strategist agent. When it completes, product-concept.md exists in the vision directory.
Step 2.5: Gate G-Stress — Stress Test Decision
Show progress:
Spark Pipeline: discover --> STRESS-TEST --> naming --> arch-vision --> use-cases --> scaffold --> spike --> visual-sketch --> style-explore --> prototypes --> feature-extract
^^^^^^^^^^^
Step 1 of gate — Explain and ask:
Briefly explain stress testing and each available test:
"Before committing to an architecture, you can stress-test the product concept to validate assumptions and identify blind spots. Four tests are available:
- Synthetic User Interview — Generates synthetic perso