Arness Spark Help

Detect the user's current position in the Arness Spark exploration pipeline, render a diagram with the active stage marked, and suggest the next command to run. Supports cross-plugin awareness — detects activity in Arness Code and Arness Infra and provides hints at the bottom of the output. This skill is strictly read-only — it never modifies files or project state.

Allowed tools: Read, Glob, Grep only. This skill MUST NOT use Write, Edit, Bash, or Task tools. This skill MUST NOT invoke any agents.

Workflow

Step 0: Multi-Plugin Awareness

1. Read the project's CLAUDE.md file.
2. Extract the ## Arness section.
3. Detect other plugin field groups:
   • code_fields_present: true if Plans directory OR Specs directory present in ## Arness
   • infra_fields_present: true if Infra plans directory OR Infra specs directory present in ## Arness
4. For each detected plugin, do a quick existence check (max 2 files):
   • Code: Check if <Specs directory> contains any FEATURE_*.md or BUGFIX_*.md, OR <Plans directory> has any subdirectories
   • Infra: Check if <Infra plans directory> has any subdirectories, OR if Dockerfile or docker-compose.yml exists in the project root
5. Record { plugin_name, has_activity: bool } for each detected plugin.
6. Proceed to Step 1. The cross-plugin hints are appended AFTER own status is rendered in Step 3.

---

Step 1: Load Configuration

1. From the ## Arness section, extract Spark fields:
   • Vision directory (e.g., .arness/vision/)
   • Use cases directory (e.g., .arness/use-cases/)
   • Prototypes directory (e.g., .arness/prototypes/)
   • Spikes directory (e.g., .arness/spikes/)
   • Visual grounding directory (e.g., .arness/visual-grounding/)
   • Reports directory (e.g., .arness/reports/)

If ## Arness is missing or no Spark fields present: Show the Spark pipeline diagram with all stages unmarked. Suggest /arn-brainstorming to start a new greenfield project. Do not attempt detection.

---

Step 2: Detect Pipeline Position

Read the pipeline reference file at ${CLAUDE_PLUGIN_ROOT}/skills/arn-spark-help/references/pipeline-map.md to load detection rules, rendering templates, and next-step tables.

Check from most advanced to least advanced — first match wins:

1. Feature backlog ready (gf-feature-extract): <vision-dir>/features/feature-backlog.md exists
2. Prototype locked (gf-prototype-lock): <prototypes-dir>/locked/LOCKED.md exists
3. Clickable prototype done (gf-clickable-proto): <prototypes-dir>/clickable/final-report.md exists
4. Static prototype done (gf-static-proto): <prototypes-dir>/static/final-report.md exists
5. Style defined (gf-style-explore): <vision-dir>/style-brief.md exists
6. Visual direction chosen (gf-visual-sketch): <vision-dir>/visual-direction.md exists
7. Spike complete (gf-spike): <vision-dir>/spike-results.md exists
8. Scaffold built (gf-scaffold): <vision-dir>/scaffold-summary.md exists
9. Use cases written (gf-use-cases): At least one UC-*.md file exists in <use-cases-dir>/
10. Architecture defined (gf-arch-vision): <vision-dir>/architecture-vision.md exists
11. Naming complete (gf-naming): <vision-dir>/naming-brief.md exists OR <reports-dir>/naming-report.md exists
12. Concept reviewed (gf-concept-review): <reports-dir>/stress-tests/concept-review-report.md exists
13. Stress tests run (gf-stress-test): Any of interview-report.md, competitive-report.md, premortem-report.md, prfaq-report.md exists in <reports-dir>/stress-tests/
14. Product discovered (gf-discover): <vision-dir>/product-concept.md exists
15. Spark initialized (gf-init): Spark fields exist in ## Arness but none of the above artifacts found

Spark complete: If the detected stage is gf-feature-extract, Spark exploration is considered complete.

---

Step 3: Render Pipeline Diagram

Read the pipeline templates from ${CLAUDE_PLUGIN_ROOT}/skills/arn-spark-help/references/pipeline-map.md.

1. Place the YOU ARE HERE marker below the detected stage.
2. Display the next-step suggestion from the next-step table.
3. Spark is a single-track pipeline (no per-project subdirectories).

Own-plugin suggestions always come first. After the pipeline diagram and next-step suggestion, append cross-plugin hints from Step 0:

• If own plugin has activity: show own status first. If other plugins have activity, append 1-2 lines at the bottom:

  Other pipelines: Development (active) — `/arn-code-help` | Infrastructure (not started)
  

• If own plugin is idle (gf-init) but other plugins have activity: show own-plugin "ready to start" suggestions FIRST:

  Ready to start: run `/arn-spark-discover` to shape your product idea, or `/arn-brainstorming` for the guided wizard.
  

  Then mention others:

  You also have active work in Code — run `/arn-code-help` for details.
  

• When Spark pipeline is complete (gf-feature-extract): suggest transition:

  Spark exploration is complete. Next steps:
  - Start developing features: `/arn-planning`
  - Deploy infrastructure: `/arn-infra-wizard`
  

Keep output concise — the user wants a quick status check, not a wall of text.

---

Step 4: Answer Follow-up Questions

After presenting the pipeline status:

1. Remain available for follow-up questions about the Spark pipeline, skill purposes, or workflow decisions.
2. Reference ${CLAUDE_PLUGIN_ROOT}/skills/arn-spark-help/references/pipeline-map.md for answers — do not re-scan the filesystem unless the user explicitly asks to refresh the detection.
3. Keep answers brief and direct.

Common questions the user might ask:

• "What is Spark?" — Spark is the exploration phase for new projects. It takes a raw idea through structured discovery, validation, prototyping, and feature extraction before development begins.
• "How do I run the full Spark pipeline?" — Run /arn-brainstorming, which walks through the entire exploration pipeline with guided decision gates.
• "What are stress tests?" — Four adversarial lenses that test your product concept: synthetic user interviews, competitive gap analysis, pre-mortem failure analysis, and PR/FAQ stress testing.
• "What is naming?" — A structured 4-step brand naming methodology: strategic foundation, creative generation, qualitative scoring, and due diligence with WHOIS domain checking.
• "Can I skip stress tests?" — Yes, all stages are optional. The brainstorming wizard presents skip options at each gate.
• "How do I transition to development?" — After feature extraction, run /arn-planning in the project. Arness Code sets up automatically on first use.
• "Where is the development pipeline status?" — Run /arn-code-help for the Arness Code development pipeline.
• "Where is infrastructure status?" — Run /arn-infra-help for the Arness Infra infrastructure pipeline.
• "What are the entry points?" — /arn-brainstorming is the main Spark entry point. For development: /arn-planning. For infrastructure: /arn-infra-wizard.
• "What is dev-setup?" — /arn-spark-dev-setup is a standalone skill for defining or onboarding to a development environment (dev containers, Docker, CI workflows). It can be run at any time after greenfield config exists.
• "What is visual-strategy?" — /arn-spark-visual-strategy sets up layered visual regression testing infrastructure (capture scripts, baselines, comparison pipelines). Requires scaffold and at least style-explore to be complete.
• "What is visual-readiness?" — /arn-spark-visual-readiness validates and activates deferred visual testing layers after project milestones. Requires a visual strategy to be defined first.

---

Error Handling

• ## Arness section not found in CLAUDE.md — Show the Spark pipeline diagram with no stage marked. Suggest running /arn-brainstorming to start a new greenfield project.
• CLAUDE.md does not exist — Same as ## Arness not found.
• Spark fields present but directories don't exist — Treat as gf-init stage. The directories may not have been created yet.
• Pipeline reference file not found — Inform the user that the pipeline-map.md reference is missing and suggest reinstalling or updating the Arness Spark plugin.
• Reports directory not configured — Skip stress-test, concept-review, and naming detection (these stages require the Reports directory). Proceed with detection from gf-arch-vision downward.

Constraints

• This skill MUST NOT write or modify any files.
• This skill MUST NOT invoke any agents.
• All tools used must be read-only: Read, Glob, Grep.
• All command suggestions must use bare skill names (e.g., /arn-spark-discover, /arn-code-help). Do not use the fully qualified plugin:skill format in user-facing text.