| name | tips-selection |
| description | Interactive TIPS (Trend-Implications-Possibilities-Solutions) candidate generation workflow for smarter-service research projects. Generates 60 trend candidates across 4 dimensions and 3 horizons using web research and training knowledge, presents them for user review, and auto-selects all 60 for downstream dimension-planner integration. This skill is a mandatory prerequisite for dimension-planner when research_type is smarter-service. Use when: (1) starting smarter-service research that needs TIPS candidates, (2) user mentions "TIPS selection", "trend candidates", or wants to generate research candidates, (3) dimension-planner halts due to missing agreed-trend-candidates.json, (4) user wants to customize or regenerate trend candidates before research planning.
|
TIPS Selection
Generate and finalize 60 TIPS trend candidates for smarter-service research projects.
Why This Skill Exists
The dimension-planner skill needs a curated set of trend candidates to generate refined research questions. This skill produces those candidates by combining live web research with training knowledge, organized across 4 dimensions and 3 horizons. All 60 candidates are auto-selected — the user reviews them but doesn't need to down-select.
Prerequisites
- Research project with
research_type: smarter-service in question frontmatter
industry_sector field in question frontmatter (or extractable from research_context)
References
Read references only when entering the corresponding phase:
Workflow
Phase 0 → Phase 0.5 → Phase 1 → Phase 2 → Phase 3
│ │ │ │ │
│ │ │ │ └─ Write agreed JSON, done
│ │ │ └─ Write trend-candidates.md for user review
│ │ └─ Generate 60 candidates (web + training mix)
│ └─ Web search for trend signals (8 searches)
└─ Load project context, validate prerequisites
Track progress using TodoWrite:
- Phase 0: Initialize & Load Context
- Phase 0.5: Web Research (if enabled)
- Phase 1: Generate 60 Candidates
- Phase 2: Present Candidates
- Phase 3: Finalize Agreed Candidates
Phase 0: Initialize
Read phase-0-initialize.md.
Extract PROJECT_PATH, INDUSTRY_SECTOR, and WEB_RESEARCH_ENABLED from the question file. Validate the project is smarter-service type. If trend-candidates.md already exists with status agreed, nothing to do.
Phase 0.5: Web Research
Read phase-0.5-web-research.md.
Run 8 web searches (4 dimensions x 2 regions: global + DACH) to gather current trend signals. These signals enrich candidate generation in Phase 1. If all searches fail, proceed with training knowledge only.
Web research is enabled by default. Disable via web_research: false in question frontmatter.
Phase 1: Generate Candidates
Read phase-1-generate.md.
Generate 60 trend candidates: 5 per cell across a 4x3 matrix (4 dimensions x 3 horizons). Mix web-sourced and training-sourced candidates. Each candidate includes trend name, 3 keywords, rationale, source type, and freshness indicator.
Phase 2: Present Candidates
Read phase-2-present.md.
Write trend-candidates.md to {PROJECT_PATH}/02-refined-questions/data/. This file is a human-readable record of all 60 candidates organized by dimension and horizon. Inform the user the file is ready for review.
Phase 3: Finalize
Read phase-3-finalize.md.
Auto-select all 60 candidates and write agreed-trend-candidates.json to .metadata/. Update trend-candidates.md status to agreed. The dimension-planner skill will pick up this JSON file automatically.
Dimension Matrix
| Dimension | German | TIPS Focus | Description |
|---|
| externe-effekte | Externe Effekte | Trend (T) | External forces, regulations, market shifts |
| neue-horizonte | Neue Horizonte | Possibilities (P) | Strategic options, business model evolution |
| digitale-wertetreiber | Digitale Wertetreiber | Implications (I) | Value creation, digital impact |
| digitales-fundament | Digitales Fundament | Solutions (S) | Capabilities, infrastructure, enablers |
Horizons
| Horizon | Timeframe | Character |
|---|
| act | 0-2 years | Immediate, validated, ready for implementation |
| plan | 2-5 years | Emerging, requires preparation |
| observe | 5+ years | Future, speculative, monitoring stage |
Output Files
trend-candidates.md
Location: {PROJECT_PATH}/02-refined-questions/data/trend-candidates.md
Human-readable record with all 60 candidates in tables, organized by dimension and horizon. Includes source provenance and freshness indicators.
agreed-trend-candidates.json
Location: {PROJECT_PATH}/.metadata/agreed-trend-candidates.json
{
"metadata": {
"industry_sector": "manufacturing",
"agreed_at": "2026-03-04T10:30:00Z",
"total_candidates": 60,
"source_skill": "tips-selection",
"web_research_status": "success",
"web_sourced_count": 28,
"training_sourced_count": 32
},
"candidates": [
{
"dimension": "externe-effekte",
"horizon": "act",
"sequence": 1,
"trend_name": "EU AI Act Compliance",
"keywords": ["ai-act", "regulation", "2026"],
"rationale": "Immediate deadline pressure",
"source": "web-signal",
"source_url": "https://...",
"freshness_date": "2026-01"
}
]
}
Integration
After this skill completes, the user runs dimension-planner. That skill checks for .metadata/agreed-trend-candidates.json — if present and valid (60 candidates), it uses them; if missing, it halts with an instruction to run tips-selection first.
Error Handling
| Scenario | Response |
|---|
| Missing question file | Exit — cannot proceed without project context |
| research_type not smarter-service | Exit — this skill only handles smarter-service |
| industry_sector not found | Ask the user to provide it |
| All web searches fail | Proceed with training-only generation (warning logged) |
| trend-candidates.md already agreed | Nothing to do — inform user |
