ds-init

name	ds-init
description	[PIPELINE] Scope a new dashboard project with interactive Q&A. Creates workspace under ./.splunk-dashboard-studio/<project>/ with requirements.md. For quick dashboard generation, skip this and go directly to ds-couture + ds-create.

ds-init — Splunk Dashboard scoping

When to use

When the user says they want to build a new Splunk Dashboard Studio dashboard and no workspace exists yet (or they pass --resume).

What it does

1. Asks the user a ten-question scoping flow (below).
2. Creates ./.splunk-dashboard-studio/<project-name>/ with state.json.
3. Writes requirements.md under the workspace.
4. Prints the next recommended skill (ds-data-explore or ds-mock).

Flags

• --autopilot — skip all questions, use defaults, and let downstream skills also run with defaults.
• --quick — ask only questions 3 (goal) and 7 (has-data). Skip design phase later.
• --resume — if a workspace already exists for the project name, load it instead of re-asking.

Question flow

Ask these one at a time. Prefer multiple-choice where options are listed.

Group 1 — Context

1. Role? Options: SOC analyst / DevOps / Platform admin / Developer / Business / Other.
2. Audience? Options: Self / Team / Leadership / External.
3. Primary goal (one sentence)? Free text.

Group 2 — Content

4. Focus? Options: Technical/operational / Business/executive / Mixed.
5. Which 3–5 questions should the dashboard answer? Free text, one per line.
6. Does a similar dashboard already exist? If yes, capture the path or URL.

Group 3 — Data

7. Is the data already in Splunk? Options: yes / no / partial.
8. (if yes or partial) Which indexes/sourcetypes are relevant? Free text or comma-separated list.

Group 4 — Scope

9. Customization level? Options: standard (apply archetype conventions verbatim) / moderate (customize layout and viz choices) / bespoke (fully custom layout, heavy tailoring).
10. Nice-to-haves? Multi-select: drilldowns / alerts / scheduled reports / tokens / dark theme / other.

Defaults used by --autopilot

Field	Default
role	"Developer" (auto-detect from CLAUDE.md if mentioned)
audience	"Self"
focus	"Mixed"
questions	infer from goal sentence
reference_dashboard	null
has_data	"no" (routes to ds-mock) unless Splunk MCP is configured
indexes	[]
customization	"moderate"
nice_to_haves	["drilldowns", "tokens"]

How to produce the workspace

Once all answers are collected, assemble them into a JSON payload and invoke:

PYTHONPATH=<path-to-repo>/plugins/splunk-dashboard-studio/src python3 -m splunk_dashboards.requirements from-json - <<'JSON'
{
  "project": "<kebab-case-project-name>",
  "goal": "<goal sentence>",
  "role": "<role>",
  "audience": "<audience>",
  "focus": "<focus>",
  "questions": ["<q1>", "<q2>"],
  "has_data": "yes|no|partial",
  "indexes": ["<idx1>"],
  "customization": "standard|moderate|bespoke",
  "nice_to_haves": ["drilldowns"],
  "reference_dashboard": null,
  "autopilot": false
}
JSON

The CLI writes .splunk-dashboard-studio/<project>/state.json and .splunk-dashboard-studio/<project>/requirements.md in the current working directory.

Next step

Read the ## Next step line at the bottom of requirements.md. Invoke the suggested skill(s).

Pipeline reading list

ds-init is the entry point — every downstream stage requires specific content skills before generating real JSON. The pipeline:

ds-init → ds-data-explore / ds-mock → ds-design → ds-create → ds-validate → ds-polish → ds-deploy

Stage	Must consult
ds-design	ds-ref-design-principles (archetype + palette) + ds-pick-viz (viz selection).
ds-create	ds-viz-<type> for every viz in the layout + ds-ref-syntax + ds-ref-pitfalls final pass.
ds-validate	ds-ref-pitfalls for trap-aware manual review.
ds-polish	ds-ref-design-principles (catalog) + ds-ref-pitfalls (configuration risks).

Skipping these lookups produces dashboards that fail schema validation, render empty, or fail the Slop Test.