Generate a strategic TIPS trend report organized around investment themes (Handlungsfelder) with inline citations and verifiable claims. The user selects a report-level narrative arc from cogni-narrative's 7 story arcs (corporate-visions, technology-futures, competitive-intelligence, strategic-foresight, industry-transformation, trend-panorama, theme-thesis) — the arc frames the executive summary, bridge paragraphs between themes, and a synthesis closing section that bind investment themes into one cohesive narrative. Each investment theme internally uses the theme-thesis arc (Why Change → Why Now → Why You → Why Pay) backed by T→I→P→S value chain evidence. Reads agreed trend candidates, enriches each with web-sourced quantitative evidence via parallel agents, assembles the report with arc-framed executive summary, bridge paragraphs, theme sections, synthesis section, and claims registry. Invokes cogni-claims:claims for optional verification. Downstream polish, visualization, and export are steered via `/trends-resume`. Required pipeline: trend-scout → value-modeler → trend-report. Use when: (1) trend-scout and value-modeler have completed, (2) user wants a written trend report, (3) user mentions "trend report", "TIPS report", "write up trends", "summarize trends", "trend analysis document", "strategic stories", (4) preparing a deliverable from scouted trends, (5) user asks to "generate report from trends" or "create trend deliverable". Always use this skill when trend-scout output exists and the user wants any kind of written trend analysis — even if they don't use the exact phrase "trend report".
From cogni-trendsnpx claudepluginhub cogni-work/insight-wave --plugin cogni-trendsThis skill is limited to using the following tools:
evals/evals.jsonreferences/claims-format.mdreferences/evidence-enrichment.mdreferences/i18n/labels-de.mdreferences/i18n/labels-en.mdreferences/phase-2-strategic-themes.mdreferences/phase-3-claim-verification.mdreferences/phase-3.5-executive-polish.mdreferences/phase-5-revision.mdreferences/region-authority-sources.jsonreferences/report-arc-frames.mdreferences/report-structure.mdtrend-report-workspace/iteration-1/benchmark.jsontrend-report-workspace/iteration-1/cdo-utility-review.jsontrend-report-workspace/iteration-1/cmo-provider-review.jsontrend-report-workspace/iteration-1/eval-english-override/eval_metadata.jsontrend-report-workspace/iteration-1/eval-english-override/with_skill/grading.jsontrend-report-workspace/iteration-1/eval-english-override/with_skill/outputs/tips-project.jsontrend-report-workspace/iteration-1/eval-english-override/with_skill/outputs/tips-trend-report-claims.jsontrend-report-workspace/iteration-1/eval-english-override/with_skill/outputs/tips-trend-report.mdGuides Payload CMS config (payload.config.ts), collections, fields, hooks, access control, APIs. Debugs validation errors, security, relationships, queries, transactions, hook behavior.
Designs, audits, and improves analytics tracking systems using Signal Quality Index for reliable, decision-ready data in marketing, product, and growth.
Enforces A/B test setup with gates for hypothesis locking, metrics definition, sample size calculation, assumptions checks, and execution readiness before implementation.
Generate a strategic TIPS trend report from agreed trend-scout candidates. Organizes the report around investment themes — each investment theme tells a CxO-level story backed by T→I→P value chain evidence. Dispatches 4 parallel agents to enrich trends with web-sourced quantitative evidence, then assembles a theme-first strategic report with executive summary and claims registry.
Transform agreed trend-scout candidates into a strategic, evidence-backed report:
cogni-claims:claims/trends-resumeFull German and English support. This skill follows the shared language resolution pattern — see $CLAUDE_PLUGIN_ROOT/references/language-resolution.md.
Two language concepts:
.workspace-config.json language setting. All Phase 0 prompts, status messages, and error messages use this language.project_language, (2) workspace language, (3) en. Always ask the user to confirm or override at the start of Phase 0.Report prose, section headers, and TIPS labels all adapt to the chosen output language. Web searches run bilingually for maximum coverage. German text uses proper umlauts (never ASCII transliterations).
trend-scout completed with execution.workflow_state == "agreed" and 60 candidatesvalue-modeler completed with tips-value-model.json containing investment themescogni-narrative plugin for theme-thesis arc guidance (graceful fallback if absent — investment themes use flat structure)cogni-copywriting plugin for executive polish (graceful fallback if absent)cogni-claims plugin for claim verificationThis skill reads ALL required state from project files — it does not depend on prior conversation context. The trends-resume dashboard, earlier questions, and any preceding chat are not inputs to the report pipeline. This means context compaction is safe and recommended before starting.
Before executing Phase 0, run /compact to free working memory. Phase 2 delegates investment theme section writing to parallel agents (reducing orchestrator context from ~69% to ~25-35%), but the orchestrator still reads the value model and claims files for assembly sections. Compacting early ensures headroom.
If /compact is unavailable (e.g., non-interactive mode), proceed without it — Phase 2's agent-based architecture is designed to stay within context limits.
| Variable | Purpose | Example |
|---|---|---|
CLAUDE_PLUGIN_ROOT | Plugin installation (skills, references) | ~/.claude/plugins/marketplaces/cogni-trends |
PROJECT_AGENTS_OPS_ROOT | Workspace root where projects live (optional, set by cogni-workspace) | User's workspace directory |
CLAUDE_PLUGIN_ROOT is injected automatically from settings.local.json. PROJECT_AGENTS_OPS_ROOT is set by cogni-workspace — if not present, scripts fall back to $PWD.
This skill is a pure orchestrator. All file I/O uses Read/Write tools; web research is delegated to agents. The only shell commands needed are:
cat file1 file2 ... > output — concatenation of log files into the final reportrm -f pattern — cleanup of stale output files on re-run[ -f file ] — existence checks before concatenationAvoid jq, sed, awk, or grep for data processing — parsing JSON through the LLM keeps the workflow self-contained and avoids dependency issues.
Read references only when needed for the specific phase:
| Reference | Read when... |
|---|---|
| $CLAUDE_PLUGIN_ROOT/references/language-resolution.md | Language detection and resolution pattern |
| $CLAUDE_PLUGIN_ROOT/references/data-model.md | Understanding entity schemas and project structure |
| references/report-arc-frames.md | Arc-specific framing templates for exec summary, bridges, synthesis (Phase 0.4b + Phase 2) |
| references/phase-2-strategic-themes.md | Assembling the investment theme report (Phase 2) |
| references/report-structure.md | Dimension section templates (written by Phase 1 agents) |
| references/evidence-enrichment.md | Configuring agent web search strategy (Phase 1) |
| references/claims-format.md | Extracting/merging claims (Phase 1-2) |
| references/i18n/labels-en.md | English report headings and labels |
| references/i18n/labels-de.md | German report headings and labels |
| references/phase-3-claim-verification.md | Running claim verification (Phase 3) |
| references/phase-3.5-executive-polish.md | Polishing report prose (Phase 3.5) |
| references/phase-5-revision.md | Post-verification report revision (Phase 5) |
Track progress through these phases as you go:
Phase 0 → Phase 0.5 → Phase 1 → Phase 2 → Phase 2.5 → Phase 3 → Phase 3.5 → Phase 4 → Phase 5
│ │ │ │ │ │ │ │ │
│ │ │ │ │ │ │ │ └─ Post-verification revision
│ │ │ │ │ │ │ └─ Update metadata, recommend /trends-resume
│ │ │ │ │ │ └─ Optional executive polish
│ │ │ │ │ └─ Optional claims verification
│ │ │ │ └─ Structural review (cross-theme quality gate)
│ │ │ └─ Theme narratives + exec summary + bridges + synthesis
│ │ └─ 4 parallel agents: enrich trends, write sections, extract claims
│ └─ Optional deep research for 3-5 high-value ACT-horizon trends
└─ Project discovery, arc selection, load inputs, validate gate
Read $CLAUDE_PLUGIN_ROOT/references/language-resolution.md. Detect workspace language from .workspace-config.json (via ${PROJECT_AGENTS_OPS_ROOT}/.workspace-config.json or CWD). Set INTERACTION_LANGUAGE — use this for all user-facing messages, prompts, and status updates from this point on.
Trend-scout projects use
trend-scout-output.json(notsprint-log.json), so the sharedproject-picker.mdpattern does not apply.
--project-path was provided as argument, use it directlydiscover-projects.sh --json to enumerate all projects{path}/.metadata/trend-scout-output.json existsexecution.workflow_state == "agreed" and tips_candidates.total >= 60REQUIRED (validate only — do NOT hold candidates or signals in context):
{PROJECT_PATH}/.metadata/trend-scout-output.json
→ Extract: config.industry, config.research_topic
→ Extract: config.market_region (default: "dach" if absent — older projects pre-regionalization)
→ Extract: project_language (top-level, NOT config.language)
→ Validate: tips_candidates.total >= 60, execution.workflow_state == "agreed"
→ Do NOT extract tips_candidates.items — agents read these themselves
REQUIRED (value model — keep in context for Phase 2):
{PROJECT_PATH}/tips-value-model.json
→ Check: investment_themes[] array exists and has ≥1 entry
→ Extract: investment_themes[], value_chains[], solution_templates[]
NOTE: Raw web signals (web-research-raw.json) are NOT loaded by the orchestrator.
Phase 1 agents read and filter signals themselves — see trend-report-writer agent.
Display to the user: "{PHASE_0_INVESTMENT_THEMES_FOUND}" (from i18n labels)
The full tips-value-model.json contains scoring matrices, blueprints, and reanchor logs that Phase 2 does not need. To reduce context pressure, extract only the fields Phase 2 uses and write a pruned subset.
Read tips-value-model.json (already loaded in Step 0.2). Write {PROJECT_PATH}/.logs/phase2-value-model.json containing ONLY these top-level keys:
{
"investment_themes": [],
"value_chains": [],
"orphan_candidates": [],
"coverage": {},
"mece_validation": {},
"solution_templates": [
{ "st_id": "...", "name": "...", "category": "...", "enabler_type": "...", "investment_theme_ref": "...", "portfolio_grounding": [...] }
]
}
investment_themes, value_chains, orphan_candidates, coverage, mece_validation in fullsolution_templates[] entry, keep ONLY: st_id, name, category, enabler_type, investment_theme_ref, portfolio_grounding — omit solution_blueprint, description, and all other fields. portfolio_grounding is needed for Phase 2 portfolio close (product names and links)reanchor_log, solution_process_improvements, metrics, collaterals, portfolio_gaps, etc.)| Check | Condition | On Failure |
|---|---|---|
| Output exists | .metadata/trend-scout-output.json | HALT: Run trend-scout first |
| Workflow state | == "agreed" | HALT: Complete trend-scout selection |
| Candidate count | >= 60 | HALT: Expected 60 agreed candidates |
| Value model exists | tips-value-model.json with investment_themes[] | HALT: Run value-modeler first |
| Config complete | industry, subsector, language present | HALT: Incomplete config |
The project_language from trend-scout-output.json is the default (falling back to workspace language if not set). Always confirm with the user. Present the question in the INTERACTION_LANGUAGE:
If INTERACTION_LANGUAGE == "de":
AskUserQuestion:
question: "In welcher Sprache soll der Report erstellt werden? trend-scout hat '{project_language}' verwendet."
header: "Report-Sprache"
options:
- label: "Deutsch (DE) ← Standard"
- label: "English (EN)"
If INTERACTION_LANGUAGE == "en":
AskUserQuestion:
question: "Report language? trend-scout used '{project_language}'. Keep or change?"
header: "Report language"
options:
- label: "English (EN) ← Default"
- label: "Deutsch (DE)"
The option matching the current default gets the arrow marker. Set LANGUAGE to the user's choice. Update project_language in trend-scout-output.json if changed.
Read references/report-arc-frames.md for the full arc frame definitions.
The report-level arc determines how investment themes connect into one cohesive narrative — the executive summary voice, bridge paragraphs between themes, and the synthesis closing section. It does NOT change how individual theme sections are written internally (those always use the theme-thesis arc).
Present the 7 available arcs via AskUserQuestion. The recommended default is corporate-visions (the proven B2B persuasion frame). Auto-detect a different recommendation if the topic strongly signals another arc (e.g., heavily regulatory topics → industry-transformation).
If INTERACTION_LANGUAGE == "de":
AskUserQuestion:
question: "{PHASE_0_ARC_QUESTION}"
header: "{PHASE_0_ARC_HEADER}"
options:
- label: "{ARC_CORPORATE_VISIONS}"
description: "{ARC_CORPORATE_VISIONS_DESC}"
- label: "{ARC_TECHNOLOGY_FUTURES}"
description: "{ARC_TECHNOLOGY_FUTURES_DESC}"
- label: "{ARC_INDUSTRY_TRANSFORMATION}"
description: "{ARC_INDUSTRY_TRANSFORMATION_DESC}"
- label: "{ARC_STRATEGIC_FORESIGHT}"
description: "{ARC_STRATEGIC_FORESIGHT_DESC}"
If INTERACTION_LANGUAGE == "en":
AskUserQuestion:
question: "{PHASE_0_ARC_QUESTION}"
header: "{PHASE_0_ARC_HEADER}"
options:
- label: "{ARC_CORPORATE_VISIONS}"
description: "{ARC_CORPORATE_VISIONS_DESC}"
- label: "{ARC_TECHNOLOGY_FUTURES}"
description: "{ARC_TECHNOLOGY_FUTURES_DESC}"
- label: "{ARC_INDUSTRY_TRANSFORMATION}"
description: "{ARC_INDUSTRY_TRANSFORMATION_DESC}"
- label: "{ARC_STRATEGIC_FORESIGHT}"
description: "{ARC_STRATEGIC_FORESIGHT_DESC}"
AskUserQuestion limit: The picker supports max 4 options. Present the 4 most relevant arcs based on the topic. The recommended arc is always first (with arrow marker). If the user selects "Other", show the remaining 3 arcs (
competitive-intelligence,trend-panorama,theme-thesis) in a follow-up question.
Set REPORT_ARC_ID to the user's choice (e.g., corporate-visions, technology-futures). This variable is passed to Phase 2 for arc-aware assembly.
The research question ({TOPIC}) becomes the subtitle. Generate a punchy title (max 8 words) from {TOPIC}, REPORT_ARC_ID, INDUSTRY, and investment theme names (already loaded in Step 0.2). The title should:
The subtitle is {TOPIC}, optionally shortened for readability (drop redundant geographic/industry qualifiers if obvious from context).
Present via AskUserQuestion:
If INTERACTION_LANGUAGE == "de":
AskUserQuestion:
question: "{PHASE_0_TITLE_QUESTION}\n\n**Titel:** {proposed_title}\n**Untertitel:** {proposed_subtitle}"
header: "{PHASE_0_TITLE_HEADER}"
options:
- label: "{PHASE_0_TITLE_ACCEPT}"
- label: "{PHASE_0_TITLE_EDIT}"
If INTERACTION_LANGUAGE == "en":
AskUserQuestion:
question: "{PHASE_0_TITLE_QUESTION}\n\n**Title:** {proposed_title}\n**Subtitle:** {proposed_subtitle}"
header: "{PHASE_0_TITLE_HEADER}"
options:
- label: "{PHASE_0_TITLE_ACCEPT}"
- label: "{PHASE_0_TITLE_EDIT}"
If the user selects Edit, ask them to provide their preferred title and/or subtitle as free text. Store the final values as {TITLE} and {SUBTITLE}.
Read the labels file matching the chosen language:
On re-runs, remove stale files to prevent mixing old and new content:
rm -f "{PROJECT_PATH}/.logs/report-header.md" \
"{PROJECT_PATH}/.logs/report-section-"*.md \
"{PROJECT_PATH}/.logs/report-investment-theme-"*.md \
"{PROJECT_PATH}/.logs/report-bridge-"*.md \
"{PROJECT_PATH}/.logs/report-synthesis.md" \
"{PROJECT_PATH}/.logs/enriched-trends-"*.json \
"{PROJECT_PATH}/.logs/claims-"*.json \
"{PROJECT_PATH}/.logs/report-claims-registry.md" \
"{PROJECT_PATH}/tips-trend-report.md" \
"{PROJECT_PATH}/.logs/phase2-value-model.json" \
"{PROJECT_PATH}/tips-trend-report-claims.json" \
"{PROJECT_PATH}/tips-insight-summary.md"
When to run: Offer this phase when the user wants deeper evidence for high-value trends. Skip if the user wants a fast report or explicitly declines.
Purpose: Select 3-5 high-value ACT-horizon trends for recursive deep research before standard evidence enrichment. Deep-researched trends get richer evidence (quantitative data, forcing functions, ROI figures) that makes the Why Change / Why Pay investment theme arguments substantially more credible.
Step 0.5.1: Select Trends for Deep Research
Ask the user via AskUserQuestion:
EN: "I can perform deep research on 3-5 high-value trends before writing the report. This adds ~5-10 minutes but produces richer evidence with quantitative data. Would you like to:
a) Deep research top ACT-horizon trends (recommended for executive audiences)
b) Skip deep research and proceed with standard evidence enrichment
c) Select specific trends for deep research"
DE: "Ich kann eine Tiefenrecherche für 3-5 hochwertige Trends durchführen, bevor der Bericht geschrieben wird. Das dauert ~5-10 Minuten länger, liefert aber reichere Evidenz mit quantitativen Daten. Möchten Sie:
a) Tiefenrecherche der wichtigsten ACT-Horizont-Trends (empfohlen für Führungskräfte-Publikum)
b) Tiefenrecherche überspringen und mit Standard-Evidenzanreicherung fortfahren
c) Spezifische Trends für Tiefenrecherche auswählen"
Step 0.5.2: Auto-Select or User-Select
If user chose (a), auto-select using these criteria (in priority order):
signal_intensity >= 4 and confidence_tier == "high" — sorted by composite score descendingconfidence_tier == "medium" and highest scoresIf user chose (c), present the ACT-horizon trend list and let them pick.
Step 0.5.3: Dispatch Deep Researchers (Parallel)
Dispatch one trend-deep-researcher agent per selected trend, all in parallel:
Task:
subagent_type: "cogni-trends:trend-deep-researcher"
description: "Deep research: {TREND_NAME}"
prompt: |
Perform deep research on this trend candidate.
PROJECT_PATH: {{PROJECT_PATH}}
TREND_NAME: {{TREND_NAME}}
TREND_KEYWORDS: {{KEYWORDS}}
DIMENSION: {{DIMENSION}}
HORIZON: act
SUBSECTOR_EN: {{SUBSECTOR_EN}}
SUBSECTOR_DE: {{SUBSECTOR_DE}}
RESEARCH_HINT: {{RESEARCH_HINT}}
MARKET_REGION: {{MARKET_REGION}}
Process results: Each agent writes a .logs/deep-research-{slug}.json artifact. Log success/failure counts. These artifacts are consumed by Phase 1 trend-report-writer agents — trends with deep research artifacts skip their own WebSearch and use the richer findings directly.
Read references/evidence-enrichment.md for web search strategy. Read references/claims-format.md for claims extraction schema.
Deep research integration: Before each trend-report-writer agent runs its evidence enrichment, it checks for {PROJECT_PATH}/.logs/deep-research-{trend-slug}.json. If a deep research artifact exists for a trend in its dimension, the writer uses the artifact's synthesis and sources instead of running its own WebSearch for that trend. This is a fourth evidence status alongside signal_sufficient, signal_partial, and signal_none: deep_research_available — richest evidence tier, no additional search needed.
Dispatch all 4 agents in a single message (parallel tool calls) so they run concurrently:
Per agent:
subagent_type: "cogni-trends:trend-report-writer"
model: sonnet
prompt: |
Dimension: {DIMENSION}
TIPS Role: {TIPS_ROLE}
Project Path: {PROJECT_PATH}
Language: {LANGUAGE}
Market Region: {MARKET_REGION}
Industry EN/DE: {INDUSTRY_EN} / {INDUSTRY_DE}
Subsector EN/DE: {SUBSECTOR_EN} / {SUBSECTOR_DE}
Topic: {TOPIC}
Labels: {relevant i18n labels}
Agents self-load candidates and raw signals from disk using PROJECT_PATH — no need to pass data in the prompt. This keeps the orchestrator context lean for Phase 2.
Dimensions: externe-effekte (T), digitale-wertetreiber (I), neue-horizonte (P), digitales-fundament (S).
Each agent writes:
{PROJECT_PATH}/.logs/report-section-{dimension}.md — narrative section (dimension-level prose){PROJECT_PATH}/.logs/claims-{dimension}.json — extracted claims{PROJECT_PATH}/.logs/enriched-trends-{dimension}.json — per-trend evidence blocks keyed by candidate_ref; actions_md uses semicolon-separated keywords (used in investment theme assembly)Each agent returns compact JSON with ok, dimension, trends_covered, claims_extracted, signal reuse metrics, and the three output file paths (section_file, claims_file, enriched_file).
If an agent returns ok: false: retry once. If retry also fails: HALT with the dimension name. All 4 must succeed before Phase 2.
After all 4 agents complete, verify that all 12 expected files exist:
For each dimension in [externe-effekte, digitale-wertetreiber, neue-horizonte, digitales-fundament]:
✓ {PROJECT_PATH}/.logs/report-section-{dimension}.md — narrative section (intermediate artifact)
✓ {PROJECT_PATH}/.logs/claims-{dimension}.json — extracted claims
✓ {PROJECT_PATH}/.logs/enriched-trends-{dimension}.json — per-trend evidence blocks (required for Phase 2)
If any report-section-{dimension}.md file is missing, log a WARNING. Phase 2 can proceed (it uses enriched-trends).
CRITICAL: You MUST read references/phase-2-strategic-themes.md before starting Phase 2. The report is organized by investment themes from tips-value-model.json, NOT by TIPS dimension. Do NOT simply concatenate the dimension section files from Phase 1 — those are intermediate artifacts, not the final report structure.
EXECUTION — Agent-assisted investment theme writing + orchestrator assembly. Phase 2 delegates the context-heavy investment theme section writing to parallel trend-report-investment-theme-writer agents (one per investment theme), then the orchestrator writes the remaining lightweight sections (executive summary, claims registry) and concatenates the final report.
Summary of steps (details in the reference):
.logs/phase2-value-model.json for investment themes, value chains, solution templates, orphan candidates, coverage datacogni-trends:trend-report-investment-theme-writer agent with MARKET_REGION: {MARKET_REGION} and REPORT_ARC_ID: {REPORT_ARC_ID} in the prompt. All agents in a single message (parallel). Each agent self-loads evidence from disk, writes report-investment-theme-{investment_theme_id}.md, and returns compact JSON with word count, citation count, quality gate status, and top claims.ok: true and quality gates passed. Retry once on failure.report-investment-theme-{investment_theme_id}.md files. Use REPORT_ARC_ID to select the arc-specific opener and closer patterns from report-arc-frames.md. Write report-header.md.report-bridge-{N}-{N+1}.md files.report-synthesis.md.claims-{dimension}.json files once, map claims to investment themes via value model, write report-claims-registry.mdtips-trend-report.mdtips-trend-report-claims.jsonResume logic: Before dispatching an agent for an investment theme, check if report-investment-theme-{investment_theme_id}.md already exists and is >1000 bytes. If so, skip that agent — display "{PHASE_2_INVESTMENT_THEME_AGENT_SKIP_RESUME}" and continue. This means re-runs only dispatch for missing investment themes.
Purpose: Cross-theme quality check of the assembled report before claims verification. Individual agents have their own quality gates, but this phase catches issues that span themes — duplicate evidence, inconsistent forcing functions, themes with zero quantitative data, missing Handeln/Nichthandeln contrasts.
Step 2.5.1: Dispatch Reviewer Agent
Task:
subagent_type: "cogni-trends:trend-report-reviewer"
description: "Structural review of trend report"
prompt: |
Review the assembled trend report for structural quality.
PROJECT_PATH: {{PROJECT_PATH}}
REPORT_PATH: {{PROJECT_PATH}}/tips-trend-report.md
REVIEW_ITERATION: 1
OUTPUT_LANGUAGE: {{PROJECT_LANGUAGE}}
Step 2.5.2: Process Verdict
If verdict is "accept" (score >= 0.80): proceed to Phase 3.
If verdict is "revise":
revision_priorities from the verdicttips-trend-report.md:
REVIEW_ITERATION: 2Skip condition: If the user explicitly asks for a fast report or says "skip review", bypass this phase.
Read references/phase-3-claim-verification.md for the full workflow. Asks the user whether to verify extracted claims via cogni-claims:claims. If the plugin is not installed, skip with a warning.
Read references/phase-3.5-executive-polish.md for the full workflow. Polishes report prose via cogni-copywriting:copywriter with SCOPE=tone. Validates citations and structure are preserved; reverts on failure.
Update {PROJECT_PATH}/tips-project.json with current timestamp:
{ "updated": "ISO-8601" }
Add to {PROJECT_PATH}/.metadata/trend-scout-output.json:
{
"trend_report_complete": true,
"trend_report_path": "tips-trend-report.md",
"trend_report_claims_path": "tips-trend-report-claims.json",
"trend_report_mode": "strategic-themes",
"trend_report_investment_theme_count": N,
"trend_report_generated_at": "ISO-8601"
}
Trend Report Complete (Investment Themes)
─────────────────────────────────────────
Report: {PROJECT_PATH}/tips-trend-report.md
Themes: {N} investment themes ({REPORT_ARC_ID} arc)
Claims: {PROJECT_PATH}/tips-trend-report-claims.json
Trends: 60 across {N} investment themes
Claims: {total_claims} quantitative claims extracted
Verification: {verdict or "skipped"}
Run /trends-resume to see your full options — polish, visualize, present, export, or accumulate.
Read references/phase-5-revision.md for the full workflow. Triggers when claims verification (Phase 3) or a later /claims session produced resolved deviations that require report corrections or claim removals. If all claims verified clean or the user skipped verification, this phase is skipped.
Dispatches the trend-report-revisor agent to:
tips-trend-report-v2.md) that becomes the new canonical reportThis phase also handles the deferred flow: when the user verified and resolved claims in a separate session via /claims and returns via /trends-resume, the resume skill detects the revision phase and routes here.
| Scenario | Action |
|---|---|
trend-scout-output.json missing | HALT: Run trend-scout first |
workflow_state != "agreed" | HALT: Complete candidate selection |
tips_candidates.total < 60 | HALT: Expected 60 candidates |
tips-value-model.json missing or no investment themes | HALT: Run value-modeler first |
tips-value-model.json has investment themes but no value chains | HALT: value-modeler Phase 1 incomplete |
| No raw signals file (both sources) | WARNING: proceed without signals (~120 searches) |
Phase 1 agent returns ok: false | Retry once, then HALT with dimension name |
| All 4 Phase 1 agents fail | HALT: Check web access is enabled |
| enriched-trends JSON missing | HALT: Phase 1 agent failed to produce enriched output |
Investment theme agent returns ok: false | Retry once, then HALT with investment theme name |
| Investment theme agent quality gate fails | WARNING: continue (section written but may be thin) |
| Investment theme references unknown candidate_ref | WARNING: agent skips that candidate in investment theme narrative |
cogni-narrative not installed | WARNING: investment-theme-writer uses flat structure (no arc guidance) |
cogni-copywriting not installed | WARNING: skip executive polish |
cogni-claims not installed | WARNING: skip verification |
| claims verification returns FAIL | Present failed claims. Do not auto-correct. |
Upstream:
trend-scout produces trend-scout-output.json (required)value-modeler produces tips-value-model.json (required)Pipeline: trend-scout → value-modeler → trend-report
Optional cross-plugin: cogni-narrative theme-thesis arc (Phase 2 investment theme writer guidance), cogni-claims:claims (Phase 3)
Downstream (via /trends-resume): cogni-copywriting:copywrite (prose polish), cogni-visual:enrich-report (themed HTML), cogni-visual:story-to-slides (presentation), cogni-visual:story-to-web (landing page), cogni-visual:story-to-big-picture (journey map), cogni-visual:story-to-storyboard (print posters), trends-catalog import, trends-dashboard
Log files in {PROJECT_PATH}/.logs/:
report-header.md — frontmatter + exec summaryreport-section-{dimension}.md — dimension sections (4 files, written by agents)phase2-value-model.json — pruned value-model subset for Phase 2enriched-trends-{dimension}.json — per-trend evidence blocks (4 files, used in investment theme assembly)report-investment-theme-{investment_theme_id}.md — investment theme sections (3-7 files, written by investment theme agents)claims-{dimension}.json — dimension claims (4 files)report-claims-registry.md — claims tableOutput files in {PROJECT_PATH}/:
tips-trend-report.md — assembled final reporttips-trend-report-claims.json — merged claims registrytips-insight-summary.md — legacy artifact (no longer generated; cleaned up on re-runs)| Issue | Check |
|---|---|
| Phase 1 agent hangs | Verify web access is enabled |
| Investment theme agent hangs | Check enriched-trends files exist in .logs/ |
| Empty claims | Check if trends have quantitative data in trend-scout output |
| Wrong language | Verify project_language in trend-scout-output.json |
| Missing sections | Check .logs/ for partial agent output |