/kai-data-dashboard — Sourced Data As An Operator Surface

Kai root note: knowledge/, harness/, and scripts/ paths in this skill live in the Kai install, not the user's project. Resolve them against the first ancestor directory of this SKILL.md that contains a knowledge/ folder (the Kai plugin root, ~/.claude/kai, or the kai-cmo-harness repo). MARKETING.md, memory/, and any output files live in the current project. If a referenced scripts/ command is not available in this install, say so, skip it, and continue with the file-based guidance — never fabricate its output.

Objective

A dashboard handoff package that a frontend, BI team, or single static HTML file can build from without anyone inventing a number: every metric defined by formula, every metric bound to a named source, every unsupported field visible as a gap. The data already exists — this skill packages it into an operator surface.

This is a companion surface. It does not replace analytics setup, audit analysis, or outbound strategy.

Done when

Work type audit-report — floor E3/C4/O1 (harness/eco-floors.yaml). This package is client- or operator-facing and carries quantitative claims, so it inherits the audit floor rather than the internal default.

• E3 — a named human approved the exact delivered files, and every number in the spec resolves to a row in source-map.md or the source folder's _data-sources.md.
• C4 — the Kai Data Provenance Rule holds. Mode declared, sources cited, gaps written down. For audit source folders, audit_provenance_lint.py passes.
• O1 — every alert threshold names the metric it watches and whether the threshold is sourced or a stated hypothesis. A threshold with no label is not finished.

Constraints

• A source is required. Accept workspace/sdr-operator/<package-slug>/, any folder holding kai-data.json, audit-data.json, _data-sources.md, or _data-gaps.md, CSV exports from CRM/ESP/sequencer/ads/analytics/sales tools, markdown reports with source-backed findings, or user-provided metrics and targets. If there is no source folder or file, ask for it.
• No fabricated sample data unless the output is explicitly labeled internal_demo.
• Declare the data mode on the artifacts: sales_external, onboarding_connected, user_provided, or internal_demo.
• Read the provenance ledgers before designing anything — _data-sources.md, _data-gaps.md, and the available JSON/CSV files. Do not add numbers that are not present in the source. Unsupported fields are listed as gaps.
• For audit source folders, run the blocking lint:

  python scripts/quality_gates/audit_provenance_lint.py <source-folder> --audit-dir
  

• Before handoff: every number has a source, retrieval date, or internal_demo label; every metric has a formula or definition; _data-gaps.md / data-gaps.md is represented in the surface; no placeholder text remains; sensitive fields are excluded, masked, or explicitly approved. HTML output is readable at desktop and mobile widths.
• Gaps are a first-class panel, never hidden or filled in.
• Do not build a full frontend app unless the user asks for implementation. For app builds, hand the spec to the relevant frontend skill or repository code.
• Static HTML rules when index.html is produced: single file unless app integration is requested; tables for dense operator data; restrained styling with readable status colors and responsive layout; critical numbers as text rather than only canvas or images; a source footer or source drawer; empty states for missing metrics.

Context

Need	Load
Provenance rule, modes, source tiers, gap handling	harness/references/audit-data-provenance.md
Source-folder ledgers and collected data	_data-sources.md, _data-gaps.md, kai-data.json / audit-data.json in the source folder
Tracking plan or attribution model instead	/kai-analytics
Client-ready audit deck instead	/kai-html-presentation
Durable, white-labeled client product with onboarding wizard and page set	/kai-client-dashboard (it calls this skill for the data-contract layer underneath it)

Dashboard type — pick from source and request. Default to sdr_operator_room when the source came from /kai-sdr-operator.

Type	Best fit	Primary view
sdr_operator_room	SDR package, lead ledger, reply data	Pipeline state, source quality, next actions
marketing_ops	Campaign, content, SEO, ad, lifecycle data	Channel performance and bottlenecks
executive_scorecard	Monthly/weekly report	KPIs, decisions, risks, next steps
audit_delivery	Audit folder	Findings, scorecards, fixes, data gaps
connector_health	API sync or integration data	Source freshness, failures, missing credentials

Spec contract. Every dashboard spec carries: audience (executive, operator, SDR, marketer, client, founder, or analyst); jobs to be done; metric definitions with exact formulas; data source per metric; refresh cadence and freshness warning; widgets, filters, drilldowns, empty states, error states; alert thresholds with source or hypothesis label; permissions and sensitive-data handling; handoff notes for frontend, BI, or static HTML build.

SDR dashboards additionally carry: status counts by sourced, enriched, approved_for_copy, queued, sent, replied, meeting_booked, disqualified, suppressed, blocked; source quality table; fit score distribution; next-action queue; reply triage categories; suppression, bounce, opt-out and complaint warnings; data gaps that block live outreach.

Output goes to <source-folder>/dashboard/: dashboard-spec.md, metrics-dictionary.md, data-contract.json, source-map.md, data-gaps.md, plus index.html when the user asks for a usable static artifact. Report the folder path, dashboard type, files produced, data gaps, and whether HTML was built or only specified.

Escalate when

• No source folder or file was given, or the named source contains no retrievable data.
• The user wants a metric the source cannot support — that is a gap, and forcing it needs a human decision.
• Sensitive fields (PII, contact records, call recordings, revenue detail) would appear on a surface whose permissions are undefined.
• The request is a full frontend app build rather than a spec plus optional static file.
• Sample data is requested for a surface that will not be labeled internal_demo.