TIPS Resume

Session entry point for returning to trend scouting work. This skill orients the user by showing where they left off and what to do next — the dashboard view that keeps multi-session TIPS projects on track.

Core Concept

TIPS projects span multiple sessions and skills (trend-scout → value-modeler → trend-report → verification). Without a clear re-entry point, users lose context between sessions and waste time figuring out what they already did. This skill bridges that gap: it reads the project state, surfaces progress at a glance, and recommends the most valuable next step.

Workflow

1. Find TIPS Projects

Discover TIPS projects in the workspace using the discovery script.

Determine the workspace root before calling the script. The workspace root is the directory that contains the cogni-trends/ project folder. Check in this order:

1. $PROJECT_AGENTS_OPS_ROOT — if set (via settings.local.json env block), use it
2. Otherwise, look for a cogni-trends/ directory under the current working directory
3. If not found under $PWD, check if there's a .workspace-config.json nearby that indicates the workspace location

Pass the root explicitly with --root:

bash $CLAUDE_PLUGIN_ROOT/scripts/discover-projects.sh --json --root "<workspace-root>"

If you cannot determine a specific root, omit --root and the script falls back to $PROJECT_AGENTS_OPS_ROOT or $PWD:

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

Returns JSON with count and projects array. Each project includes path, slug, industry, subsector, research_topic, workflow_state, candidates_total, and has_report.

The script searches:

1. The workspace root (from --root, $PROJECT_AGENTS_OPS_ROOT, or $PWD) for cogni-trends/*/tips-project.json
2. The global project registry (~/.claude/cogni-trends-projects.json) for projects created in other workspaces

If count is 0:

• First, ask the user if they have a project in a different directory (e.g., OneDrive, external workspace). If they provide a path, register it:

  bash $CLAUDE_PLUGIN_ROOT/scripts/discover-projects.sh --register "<path>"
  

  Then re-run discovery.
• If no path is provided, suggest the trend-scout skill to start a new project.

2. Select Project

• One project found — use it automatically.
• Multiple projects — present them with industry + topic + status and ask which one to continue.

3. Run Project Status with Health Check

bash $CLAUDE_PLUGIN_ROOT/scripts/project-status.sh "<project-dir>" --health-check

The script returns JSON with project, counts (including blueprints, anchored_solutions, avg_readiness), portfolio_anchors (per-product breakdown with needs coverage and quality flags), scoring, artifacts, portfolio_bridge, phase, next_actions, and stale_warnings.

4. Present Status Summary

Show a concise, scannable dashboard. Lead with the project name and industry context, then the progress overview:

Project Header:

TIPS Project: {research_topic}
Industry: {industry} / {subsector}
Language: {language}

Progress Table:

Stage	Status	Details
Web Research	Done / Pending	{web_research_status}, {candidates_web} signals found
Candidate Generation	Done / Pending	60 generated
Candidate Selection	Done / Pending	{candidates_total}/60 agreed
Portfolio Bridge	Done / Ready / N/A	v{context_version} context, {features_count} features
Value Chains & Themes	Done / Pending	{themes_count} strategic themes
Solution Templates	Done / Pending	{solutions_count} solutions generated
BR Scoring & Ranking	Done / Pending	{ranked_count} solutions ranked
Solution Blueprints	Done / Pending / N/A	{blueprints}/{solutions_count} blueprinted, avg readiness {avg_readiness}, {anchored_solutions} portfolio-anchored
Portfolio Anchors	Done / N/A	{products_count} products, {features_count} features, {delivered}/{unmet} needs, {quality_issues} quality flags
Trend Report	Done / Pending	{report_sections}/4 sections
Claims Registry	Done / Pending	{claims_total} claims extracted
Insight Summary	Done / Skipped
Claim Verification	Done / Pending / Skipped	{verdict}: {passed} passed, {failed} failed
Executive Polish	Done / Skipped	tone (cogni-copywriting)
Visual Report	Done / Skipped	themed HTML with charts (cogni-visual:enrich-report)
Dashboard	Done / Skipped	interactive HTML visualization

Solution Blueprints row — derived from counts.blueprints, counts.anchored_solutions, counts.avg_readiness:

• Done: blueprints > 0 — show blueprint count, average readiness score, and anchored count
• Pending: solutions_count > 0 but blueprints = 0 — solutions exist but no blueprints generated yet
• N/A: solutions_count = 0 — no solutions generated yet

Portfolio Anchors row — derived from portfolio_anchors in status JSON:

• Done: portfolio_anchors.total > 0 — show product count, feature count, delivered/unmet needs, and quality flag count
• N/A: portfolio_anchors.total = 0 — no portfolio-anchored solutions exist

Portfolio Anchor Health (show when portfolio_anchors.total > 0, after the Progress Table):

Render a per-product summary table from portfolio_anchors.products:

Product	Features	Solutions	Delivered	Unmet	Quality
{product_slug}	{features}	{solutions}	{needs_delivered}	{needs_undelivered}	OK or {quality_issues} flags

Coverage above 70% (delivered / total needs) indicates healthy anchoring. Products with quality flags need attention before customer-facing use — point users to /trends-dashboard for per-solution detail.

Portfolio Bridge row — derived from portfolio_bridge in status JSON:

• Done: context_file is true — show version and features count
• Done (upgrade available): context_file is true but context_version < "3.1" — show current version and recommend: "Run /bridge portfolio-to-tips to upgrade to v3.1 for provider differentiators in trend-report."
• Ready: portfolio_project_found is true but context_file is false — recommend running /bridge portfolio-to-tips before value-modeler
• N/A: portfolio_project_found is false — no portfolio project in workspace

When the phase is modeling or modeling-paths and the Portfolio Bridge status is Ready, lead the next action recommendation with: "Before starting the value modeler, run /bridge portfolio-to-tips to ground your solutions in real products."

When Portfolio Bridge status is Done (upgrade available) and the phase is reporting or later, add to recommendations: "Re-run /bridge portfolio-to-tips to add provider differentiators — trend-report will use them for stronger portfolio close sections."

Scoring Summary (if candidates exist):

• Average score: {avg_score}
• Leading indicators: {leading_pct}%
• Confidence: {high} high, {medium} medium, {low} low

Dimension Balance (if candidates exist):

Dimension	ACT	PLAN	OBSERVE	Total
Externe Effekte	N	N	N	N
Neue Horizonte	N	N	N	N
Digitale Wertetreiber	N	N	N	N
Digitales Fundament	N	N	N	N

After the tables:

• Phase — translate the phase value into plain language (see reference below)
• Stale warnings — if stale_warnings is non-empty, show them as priority actions before regular next steps
• Artifacts — note which log files and outputs exist

Keep the tone warm and oriented toward action — this is a welcome-back moment, not a status report.

5. Recommend Next Action

Present each entry from next_actions with the skill name and reason. Offer to proceed with the top recommendation immediately.

If the phase is complete, congratulate the user and present the downstream options grouped by purpose (see "Downstream Options for Completed Reports" below). Highlight the top 2-3 most impactful actions based on what hasn't been done yet. Offer to proceed with the user's choice.

Phase Reference

Phase	Meaning	What to do
scouting	No candidates yet	Run trend-scout
researching	Web research in progress	Re-invoke trend-scout to continue
generating	Candidate generation in progress	Re-invoke trend-scout to continue
selecting	Candidates presented, finalizing	Re-invoke trend-scout to finalize
modeling	Candidates agreed, value model not yet built	Run /bridge portfolio-to-tips if portfolio exists, then value-modeler
modeling-paths	Relationship networks built, solutions pending	Continue value-modeler
modeling-scoring	Solutions generated, BR scoring pending	Continue value-modeler
modeling-curating	Ranked solutions complete, curation pending	Continue value-modeler for optional catalog curation
modeling-complete	Value model complete with ranked solutions	Run trend-report, or /trends-catalog import
reporting	Value model complete, report not yet generated	Run trend-report
verification	Report done, claims pending verification	Run cogni-claims:claims
revision	Claims verified and resolved, report revision pending	Run trend-report Phase 5 (revision)
complete	All stages finished	Report complete — choose from downstream options below

Stale Blueprints: When stale_warnings contains a stale_blueprints entry (portfolio context was updated after blueprints were generated), prepend a re-anchor recommendation to the next actions for phases modeling-scoring, modeling-curating, modeling-complete, reporting, and complete:

"Portfolio context has changed since blueprints were generated. Run 're-anchor solutions' via the value-modeler to update solution mappings with current portfolio data."

This is automatically handled by project-status.sh --health-check, which prepends the re-anchor action to next_actions when the condition is detected.

Downstream Options for Completed Reports

When phase is complete, the next_actions array from project-status.sh contains the full set of downstream options. Present them grouped by purpose:

Polish & Verify

• cogni-copywriting:copywrite — Polish report prose for executive readability (tone scope)
• cogni-claims:claims — Verify extracted claims against cited sources
• trend-report Phase 5 — Revise report after claims resolution (apply corrections, remove unverifiable claims)

Visualize

• cogni-visual:enrich-report — Themed HTML with Chart.js visualizations and Excalidraw diagrams
• cogni-visual:story-to-slides — PowerPoint presentation brief
• cogni-visual:story-to-web — Scrollable web landing page
• cogni-visual:story-to-big-picture — Single-canvas visual journey map
• cogni-visual:story-to-storyboard — Multi-poster print storyboard

Accumulate

• cogni-trends:trends-catalog — Import to industry catalog for cross-pursuit reuse

Dashboard

• cogni-trends:trends-dashboard — Interactive HTML dashboard of full project lifecycle

Only show actions that appear in next_actions (e.g., skip copywriting if already applied, skip enrich-report if already done, skip dashboard if already generated). Present the top 2-3 as recommended and the rest as "also available". Offer to proceed with the user's choice immediately.

All visualization skills (story-to-*) consume tips-trend-report.md directly — no intermediary step (like cogni-narrative) is needed. Pass the report path as source_path and extract arc_id from the report's YAML frontmatter to pass as the arc_id parameter — this ensures correct arc propagation even if frontmatter parsing is inconsistent.

Multi-Session Design

This skill is the recommended re-entry point after heavy sessions. TIPS work naturally spans multiple sessions — web research, candidate generation, candidate review, value modeling, and report writing each consume significant context. Other TIPS skills proactively recommend /trends-resume when they detect a heavy session (Phase 1 web research, Phase 2 generation, or report assembly completed).

When presenting the status summary, acknowledge what the user accomplished in previous sessions if recent timestamps suggest productive recent work. This continuity helps users feel their work persists and builds confidence in the multi-session workflow.

Language Support

This skill follows the shared language resolution pattern — see $CLAUDE_PLUGIN_ROOT/references/language-resolution.md.

• Communication Language: 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 or trend-scout-output.json. Communicate with the user in this language (status messages, instructions, recommendations, questions). Technical terms, skill names, and CLI commands remain in English. If no language is found anywhere, default to English.