Single-prompt deep dive

Brand report ID: [BRACKETS]
prompt: [prompt]

Execute the following task using the OtterlyAI API. Do not ask clarifying questions before starting 
— begin immediately, surface questions only if a required value is truly missing 
(e.g. country is ambiguous).
---
## STEP 1 — Resolve inputs
1. Fetch brand report metadata:
   GET /v1/reports/brand/{reportId}
   
   Extract:
   - Brand name (the main brand being tracked)
   - countries[] — if only one, use it silently; if multiple, ask the user which one
   - List of prompts to find the matching promptId by name (fuzzy match is fine)
2. If promptId is not already known, fetch the prompt list:
   GET /v1/reports/brand/{reportId}/prompts
   ?country=XX&startDate=YYYY-MM-DD&endDate=YYYY-MM-DD&limit=50
   
   Find the prompt matching the user's name. If multiple close matches, show them 
   and ask the user to confirm. Default date range: last 30 days.
---
## STEP 2 — Structured prompt analytics
Call the single-prompt detail endpoint:
GET /v1/reports/brand/{reportId}/prompts/{promptId}
?country=XX&startDate=YYYY-MM-DD&endDate=YYYY-MM-DD
Extract and present:
### 2a. Prompt identity
- Prompt text (exact wording tracked)
- Intent volume (monthly search volume estimate)
- Tags (e.g. Branded, BoFu, ToFu)
- Domain category distribution (where citations are coming from: 
  News/Media, Brand, Community/Forum, etc.)
### 2b. Brand rank & share
Build a comparison table of all detected brands:
| Brand | Rank | Mentions | Brand Coverage | Sentiment (NSS) | Positive / Neutral / Negative |
|-------|------|----------|---------------|-----------------|-------------------------------|
| Your Brand ⭐ | # | # | XX% | +XX | XX% / XX% / XX% |
| Competitor A | # | # | XX% | +XX | XX% / XX% / XX% |
| ...  |
Highlight the main brand row. Flag if rank > 3 or brand coverage < 30%.
### 2c. Trend — brand coverage history
Pull brandCoverageHistory[] for the main brand and top 1–2 competitors.
Show a simple ASCII sparkline or describe the trend clearly:
- "Coverage improved from XX% → XX% over the period"
- "Coverage dropped XX points week-over-week"
- Call out any inflection points
---
## STEP 3 — What the AI engines are actually saying
Call:
GET /v1/reports/brand/{reportId}/prompts/{promptId}/ai-responses
?country=XX&startDate=YYYY-MM-DD&endDate=YYYY-MM-DD
Show the 2 most recent runs per engine. Group output by engine. For each:
**[Engine name] — [Run date]**
- Was the brand mentioned? (Yes / No / How many times)
- Key excerpt: quote 1–2 sentences where the brand appears (or note absence)
- Top 3 citations: title + domain + rank position
- Key competitors mentioned in this response
If the brand was NOT mentioned in a response, flag it clearly: 
"⚠️ [Brand] was absent from this [Engine] response."
---
## STEP 4 — Citation sources for this prompt
Cross-reference with the citations endpoint, filtered by engines and date range, 
to identify which specific URLs are winning citations on this prompt:
GET /v1/reports/brand/{reportId}/citations
?country=XX&startDate=YYYY-MM-DD&endDate=YYYY-MM-DD&limit=20
Surface the top 5–8 most-cited domains/URLs relevant to this prompt's topic. 
For each, note:
- Domain name and category (News/Media, Brand, Community/Forum, etc.)
- Whether it's a competitor domain
- Whether the main brand's domain is cited — if not, flag the citation gap
---
## STEP 5 — Recommendations
Call:
GET /v1/reports/brand/{reportId}/recommendations?country=XX
Filter to recommendations that are relevant to this prompt's topic or category. 
Prioritize by:
1. state: suggested → todo (skip archived/removed)
2. priority: high → medium → low
3. group: on_page first, then off_page, then general
Present the top 5 actionable recommendations in this format:
**[Priority] — [Recommendation type]**
What to do: [clear, specific action]
Why it matters for this prompt: [1-sentence explanation]
Opportunity: [URL or domain if available from data.opportunities[]]
If no recommendations are prompt-specific, surface the top 3 highest-priority 
general recommendations and note they apply broadly.
---
## OUTPUT FORMAT
Structure the entire response with these sections:
---
### 🔍 Prompt Deep Dive: "[Prompt text]"
Report: [Report name] | Country: [XX] | Period: [start] → [end]
---
### 📊 Who's Winning This Prompt
[Brand rank table from Step 2b]
**Your brand's position:** [1-liner summary — winning / competitive / behind]
---
### 📈 Trend
[Coverage history narrative + sparkline from Step 2c]
---
### 🤖 What AI Engines Are Actually Saying
[Per-engine breakdown from Step 3]
---
### 🔗 Citation Sources
[Top cited URLs/domains from Step 4]
[Citation gap note if brand domain is absent]
---
### ✅ Top Recommendations to Show Up
[Top 5 actions from Step 5, formatted as above]
---
## RULES
- Default to last 30 days if no date range is specified.
- Always surface the main brand's metrics first, competitors second.
- If brand coverage < 20%, lead with that — it's the most critical signal.
- Never dump raw JSON. Translate numbers into plain-English insights.
- Round percentages to 1 decimal place. Counts > 1000 → "1.7K".
- If an AI response shows the brand absent, treat it as a high-priority finding 
  worth calling out prominently, not buried.
- If the promptId can't be resolved from the name provided, list the 10 most 
  similar prompts and ask the user to pick one.
- Cite the date range and country used in every section header.

Tip: replace anything in [BRACKETS] with your own values, then paste this into Claude with the OtterlyAI connector enabled.