TIPS Dashboard

Generate a self-contained HTML dashboard that visualizes the entire TIPS project lifecycle — from trend scouting candidates through value modeling themes to report and catalog status. The dashboard opens in the user's browser with an interactive three-panel layout inspired by knowledge-graph explorers.

Context Independence

This skill reads ALL required state from project files (trend-scout output, value model, report, catalog) — it does not depend on prior conversation context. If invoked after trends-resume or at the end of a heavy session, context compaction is safe and recommended before starting.

Before executing Step 1, run /compact to free working memory. Dashboard generation reads multiple large JSON files and produces a self-contained HTML document with embedded D3 visualizations — maximum available context ensures the full project data can be processed in one pass.

If /compact is unavailable or this is the first skill in the session, skip compaction and proceed directly.

Core Concept

TIPS projects produce rich structured data across multiple phases (trend-scout → value-modeler → trend-report → catalog). This data lives in separate JSON files and markdown outputs that are hard to reason about in aggregate. The dashboard turns all of it into a single interactive visual — navigating phases, drilling into entities, and exploring TIPS relationships (Trend → Implication → Possibility → Solution) through a force-directed graph.

It complements trends-resume (quick text-based status check) the way portfolio-dashboard complements portfolio-resume in cogni-portfolio: one is for quick re-entry, the other for visual exploration and presentation.

Information Architecture

The layout follows a three-panel design with phase-based navigation:

┌─────────────────────────────────────────────────────────────┐
│ NAVBAR: [Overview] [Scout] [Value Model] [Report] [Catalog] │
└─────────────────────────────────────────────────────────────┘
┌────────────┬──────────────────────────┬─────────────────────┐
│ LEFT       │  MAIN CONTENT            │ RIGHT PANEL         │
│ Section    │  (phase-specific)        │ ┌─────────────────┐ │
│ Index      │                          │ │   TIPS Graph    │ │
│            │                          │ │   (D3 force)    │ │
│ - item     │                          │ │   T → I → P → S │ │
│ - item     │                          │ ├─────────────────┤ │
│ - item     │                          │ │  Entity Detail  │ │
│            │                          │ │  (on click)     │ │
│            │                          │ └─────────────────┘ │
└────────────┴──────────────────────────┴─────────────────────┘

Navbar — sticky horizontal bar with phase tabs. Active tab highlighted with accent color. Hash-based routing (#overview, #scout, #model, #report, #catalog).

Left Panel — persistent section index for the active tab. In Scout: dimension list. In Value Model: investment theme list. In Report: section list. Clicking scrolls the main content. Highlights current section on scroll.

Right Panel — always visible, collapsible to icon rail. Top 60%: D3 force-directed graph showing TIPS entity relationships colored by dimension, grouped by investment theme. Bottom 40%: entity detail panel populated when clicking a graph node or entity card. Draggable resize handle between zones.

Workflow

1. Find the Active TIPS Project

Discover TIPS projects using the discovery script:

bash $CLAUDE_PLUGIN_ROOT/scripts/discover-projects.sh --json

If multiple projects exist, ask the user which one to open. Store the resolved project directory path.

2. Pick Theme

Use the cogni-workspace:pick-theme skill to let the user select a theme. The skill returns theme_path, theme_name, and theme_slug.

Skip conditions (auto-select without prompting):

• The caller already provided a theme_path
• Only one theme exists in the workspace (auto-select it)

3. Generate Design Variables

Read the selected theme.md file and produce a design-variables JSON file at <project-dir>/output/design-variables.json.

The JSON must follow the schema at $CLAUDE_PLUGIN_ROOT/skills/trends-dashboard/schemas/design-variables.schema.json. See the example at $CLAUDE_PLUGIN_ROOT/skills/trends-dashboard/examples/design-variables-cogni-work.json for the exact format.

What the LLM adds beyond a raw token extraction:

• Derives surface2 (~4% darker than surface) if not explicit in the theme
• Computes accent_muted and accent_dark variants if the theme only defines accent
• Builds a Google Fonts @import URL from the font families
• Adjusts shadow opacity for dark themes (higher opacity for light-on-dark)
• Ensures WCAG AA contrast between text and background, text_light and surface_dark
• Sets radius and shadows appropriate to the theme's visual style
• Adds dimensions color palette (4 colors for the TIPS dimensions: T, I, P, S)

Required fields: theme_name, colors (all 13 keys), status (4 keys), fonts (3 keys). Optional fields with defaults: google_fonts_import (empty), radius ("12px"), shadows (standard set).

4. Generate the Dashboard

Run the dashboard generator script:

python3 $CLAUDE_PLUGIN_ROOT/skills/trends-dashboard/scripts/generate-dashboard.py "<project-dir>" --design-variables "<project-dir>/output/design-variables.json"

The script:

• Reads tips-project.json for project metadata
• Reads .metadata/trend-scout-output.json for candidates, scoring, source integrity
• Reads tips-value-model.json for themes, value chains, solution templates, rankings
• Reads tips-trend-report-claims.json for claims registry (if present)
• Reads .metadata/value-modeler-output.json for modeling state (if present)
• Checks for tips-trend-report.md and tips-insight-summary.md existence
• Runs project-status.sh for phase and progress data
• Scans catalog directory for catalog.json (if present)
• Loads the design-variables JSON for theming
• Generates a self-contained HTML file at <project-dir>/output/trends-dashboard.html
• Returns JSON: {"status": "ok", "path": "<output-path>", "theme": "<name>"}

Legacy fallback: The script also accepts --theme <path-to-theme.md> for CI/automated runs.

5. Open in Browser

open "<project-dir>/output/trends-dashboard.html"

Tell the user the dashboard is open. If they want to refresh after making changes, just rerun the script.

Dashboard Tabs

Tab 1: Overview

Project header with name, industry, subsector, research topic, language, and last-updated timestamp. Below that:

• Phase Progress Bar — visual 8-stage pipeline (Web Research → Candidate Gen → Selection → Report → Claims → Insight → Verification → Polish) with completed/active/pending states
• Scoring Summary Cards — average score, leading indicator %, confidence distribution (high/medium/low) as donut chart
• Dimension × Horizon Heatmap — 4×3 grid (dimensions as rows, horizons ACT/PLAN/OBSERVE as columns) with candidate counts per cell, color-coded by density. Click a cell to filter the graph to that dimension+horizon
• Source Integrity — web-sourced vs training-based breakdown, corroboration rate, average scores by source type

Left index: Progress stages (clickable, scrolls to section).

Tab 2: Scout

Deep dive into the 60 trend candidates from trend-scout:

• Candidate Cards — grouped by dimension, each showing: name, statement, score badge, confidence tier, signal intensity, indicator type (leading/lagging/coincident), diffusion stage, source, freshness date
• Distribution Charts — score histogram, confidence pie, indicator type breakdown, diffusion stage funnel
• Source Analysis — training-capped count, corroboration rates, web signal score vs training score comparison

Left index: Four dimensions (Externe Effekte, Neue Horizonte, Digitale Wertetreiber, Digitales Fundament — or EN equivalents based on project language). Click to scroll to that dimension's candidates.

Graph interaction: Clicking a candidate card highlights it in the right-panel graph and shows its relationships (which implications/possibilities link to it).

Tab 3: Value Model

Strategic themes and solution ranking from value-modeler:

• Theme Cards — one per strategic theme showing: name, strategic question, executive sponsor type, narrative summary, business relevance average, ranking value. Expandable to show value chains
• Value Chain Flows — visual T→I→P flow diagrams per theme, showing how trends connect through implications to possibilities. Each node in the flow is clickable (populates entity detail)
• Solution Template Ranking — sorted table of all STs with: name, category, enabler type, theme, BR score, F1+ ranking value, readiness score, linked chains count. Color-coded rows by ranking tier. Portfolio-anchored STs show an anchor badge with their feature slug. Readiness column shows the blueprint readiness score with a colored indicator (green ≥0.8, amber ≥0.5, red <0.5)
• Solution Blueprint Panel — multi-dimensional portfolio composition view for all STs:
  • Taxonomy Radar Chart: 8-spoke radar (one per B2B ICT dimension 0-7) showing aggregate blueprint coverage across all STs that have blueprints. Each spoke = mean of role-weighted coverage values for that dimension. Coverage numeric mapping: covered=1.0, partial=0.5, gap=0.0, unknown=0.0. Role weights match the readiness formula (lead=0.5, supporting=0.3, enabling=0.2). STs without blueprints are excluded from the aggregation. Immediately shows portfolio strengths and gaps at a glance
  • Blueprint Cards per ST: within each theme section, compact blueprint visualization:

    ST-001: Predictive Quality Analytics Platform
    ┌─────────────────────────────────────────────────┐
    │ ● Lead:  6.6 AI & Analytics    ██████████ 100%  │
    │ ◐ Supp:  1.4 IoT Connectivity  █████░░░░░  50%  │
    │ ● Supp:  4.6 Cloud-Native      ██████████ 100%  │
    │ ✗ Enab:  7.2 Digital Transform  ░░░░░░░░░░   0%  │
    │                                                   │
    │ Readiness: 0.68  Span: 4 dimensions              │
    └─────────────────────────────────────────────────┘
    

  • Taxonomy Gap Heatmap: rows are dimensions (0-7), cells are individual taxonomy categories within that dimension. Only show categories referenced by at least one ST's blueprint. Color each cell by worst coverage across any ST needing that category: red = at least one ST has it as gap, amber = worst is partial (no gaps), green = all STs have it covered. Show coverage percentage per dimension row (covered cells / total cells in row). Sparse dimensions (few categories referenced) are expected — not all 57 categories will appear
  • Only appears when at least one ST has a solution_blueprint
• Portfolio Anchor Coverage — when Solution Templates have generation_mode: "portfolio-anchored", shows anchor coverage stats (anchored vs abstract STs, delivered/unmet needs), collapsible theme cards with per-ST anchor details including feature/product slugs and need pills (green = delivered, red = unmet). Quality flags are surfaced. Only appears when at least one ST is portfolio-anchored
• SPIs & Metrics — collapsible section showing Solution Process Improvements and success KPIs per theme

Left index: Theme names. Click to scroll to that theme's section.

Graph interaction: The right-panel graph shows the full TIPS network — Trends (amber), Implications (cyan), Possibilities (purple), Solutions (green) — connected by value chain links. Clicking a theme in the left index filters the graph to that theme's entities.

Tab 4: Report

Full rendered trend report content from tips-trend-report.md, plus claims data:

• Report Status — metadata cards showing report/claims/insight existence, investment theme count, industry, language
• Executive Summary — full rendered executive summary with inline citation links, investment theme overview, and cost-of-inaction framing
• Investment Themes — each theme rendered as a styled card with: strategic question (blockquote), Why Change narrative, Why Now forcing functions, Why You solution capabilities, Why Pay business case. Inline citations rendered as clickable links
• Portfolio Analysis — theme overview tables, horizon distribution, MECE validation, evidence coverage (if present in the report)
• Claims Registry — table of all claims with: statement, verification status badge (verified/unverified/deviated/resolved), source URL

Left index: Status, Summary, each investment theme by name, Claims.

Tab 5: Catalog

Industry catalog status (shown only if a catalog exists):

• Catalog Header — industry, subsector, entity counts, last updated
• Taxonomy Coverage Heatmap — 8 dimensions × categories grid (from b2b-ict-portfolio taxonomy). Green = mapped, red = gap. Coverage percentage chip
• Entity Counts — TIP entities, solution templates, SPIs, metrics, collaterals
• Pursuit History — timeline of contributions from different projects
• Cross-Pursuit Analytics — trend frequency, solution popularity, BR distribution (if multiple pursuits)

Left index: Taxonomy dimensions. Click to scroll to that dimension's categories.

Right Panel: TIPS Graph

The graph is the signature element of this dashboard. It visualizes the TIPS relationship network as a D3 force-directed graph:

• Node types (color + shape):
  • Trend (T) — amber circle
  • Implication (I) — cyan diamond
  • Possibility (P) — purple square
  • Solution (S) — green hexagon; portfolio-anchored STs show a small anchor (⚓) overlay
• Edges: Value chain links (T→I, I→P, P→S) shown as directed arrows
• Grouping: Nodes cluster by strategic theme (force-directed with theme gravity wells)
• Sizing: Node radius proportional to score or BR value
• Filters: Toggle buttons in graph-controls to show/hide each TIPS type
• Interaction:
  • Hover: highlight connected edges and nodes
  • Click: populate entity-detail-zone below with full entity data
  • Drag: reposition nodes
  • Zoom/pan: mouse wheel and drag on background

Entity Detail Zone (below graph):

• Entity name, type badge, dimension badge
• Full statement/description
• Score, confidence, source (for candidates)
• BR score, ranking value, linked chains (for solution templates)
• Solution blueprint: building blocks with role badges, taxonomy refs, coverage indicators, readiness score with progress bar. Gap blocks highlighted as "investment opportunity" (for solution templates with blueprints)
• Related entities list (clickable, navigates graph)
• Portfolio anchor data (feature slug, product slug, delivered/undelivered needs, quality flag) for portfolio-anchored solution nodes

When no entity is selected, shows a prompt: "Click a node in the graph or an entity card to see details."

Shared Pattern

This dashboard follows the design-variables pattern documented at cogni-workspace/references/design-variables-pattern.md. It uses the same 3-stage flow as portfolio-dashboard: pick-theme → LLM derives design-variables.json → generator consumes JSON.

Important Notes

• The dashboard is read-only — it shows TIPS project state, it does not modify entities
• The HTML file is fully self-contained (inline CSS + JS, D3 loaded from CDN)
• Re-running the script overwrites the previous dashboard
• The dashboard lives at output/trends-dashboard.html
• Graceful degradation: Tabs for phases that haven't been completed yet show a "Not yet available" state with a recommendation to run the relevant skill. The Overview tab always works (even with just a tips-project.json)
• Communication Language: This skill follows the shared language resolution pattern (see $CLAUDE_PLUGIN_ROOT/references/language-resolution.md). Read workspace language from .workspace-config.json (via ${PROJECT_AGENTS_OPS_ROOT}/.workspace-config.json or CWD) first. If not found, fall back to project language from tips-project.json. Communicate with the user in this language. Technical terms, skill names, and CLI commands remain in English.

Session Management

Dashboard generation is a capstone operation — it gives the user a complete visual overview and typically signals a natural pause point. After generating the dashboard, always recommend starting a fresh session for next steps:

"Dashboard ready at output/trends-dashboard.html. For next steps like [recommend from next_actions], I'd recommend starting a fresh session with /trends-resume. That picks up the current state cleanly and gives you full context for the next phase."

Use the project's communication language. Frame it as helpful advice, not a limitation.