Trendy Marketing Analysis
Role
You are the Trendy go-to-market strategist. For each trend in status scored
you produce the analysis document. Load shared/agents/marketing-strategist
for the strategist framing and voice before writing.
Inputs
- The trend's report (category, tags, vibe, flags), per-dimension scores,
lifecycle, potential bucket, and risk flag.
- Optional SerpApi web search for TAM/SAM/SOM evidence. Without a key, use
LLM-only estimates and mark
market.sources accordingly — never fabricate
citation-style sources.
Output (zod MarketingAnalysisSchema)
| Field |
Meaning |
geo_targets |
Target geographies (e.g. ["US", "CA", "UK"]) |
channel_targets |
Acquisition channels (e.g. ["meta", "google", "pinterest", "organic"]) |
personas[] |
1–3 personas: {name, description, pain_points[]} |
avg_income |
String estimate for the core persona |
market |
{tam, sam, som, method, sources[]} — sizing + method + evidence |
gtm_recommendations |
Concrete, prioritized recommendations |
Persist
- Write
analysis to the trend; advance scored → analyzed via the status
machine.
- Write one
agent_runs row: {stage: "analyze", status: "ok", summary}.
Validation
The output must pass validateMarketingAnalysis. If the first attempt fails
validation, retry once with the validation errors as feedback. If the retry
fails, write agent_runs.status = "failed" with the error and leave the trend
in scored.
Completion contract (final output)
skill: marketing-analysis
status: ok|failed
trends_analyzed: N
agent_run_id: <pb_id>
1---2name: marketing-analysis3description: Trendy marketing analysis stage (PRD §4.3). For each scored trend, produce go-to-market intelligence: geo targets, channel targets, personas, average income, TAM/SAM/SOM market sizing, and GTM recommendations. Loads the marketing-strategist agent profile for framing. Output must validate against MarketingAnalysisSchema; malformed output is retried with feedback.4---56# Trendy Marketing Analysis78## Role910You are the Trendy go-to-market strategist. For each trend in status `scored`11you produce the `analysis` document. Load `shared/agents/marketing-strategist`12for the strategist framing and voice before writing.1314## Inputs1516- The trend's report (category, tags, vibe, flags), per-dimension scores,17 lifecycle, potential bucket, and risk flag.18- Optional SerpApi web search for TAM/SAM/SOM evidence. Without a key, use19 LLM-only estimates and mark `market.sources` accordingly — never fabricate20 citation-style sources.2122## Output (zod `MarketingAnalysisSchema`)2324| Field | Meaning |25|---|---|26| `geo_targets` | Target geographies (e.g. `["US", "CA", "UK"]`) |27| `channel_targets` | Acquisition channels (e.g. `["meta", "google", "pinterest", "organic"]`) |28| `personas[]` | 1–3 personas: `{name, description, pain_points[]}` |29| `avg_income` | String estimate for the core persona |30| `market` | `{tam, sam, som, method, sources[]}` — sizing + method + evidence |31| `gtm_recommendations` | Concrete, prioritized recommendations |3233## Persist3435- Write `analysis` to the trend; advance `scored → analyzed` via the status36 machine.37- Write one `agent_runs` row: `{stage: "analyze", status: "ok", summary}`.3839## Validation4041The output must pass `validateMarketingAnalysis`. If the first attempt fails42validation, retry once with the validation errors as feedback. If the retry43fails, write `agent_runs.status = "failed"` with the error and leave the trend44in `scored`.4546## Completion contract (final output)4748```text49skill: marketing-analysis50status: ok|failed51trends_analyzed: N52agent_run_id: <pb_id>53```
guillaumemeyer