amazon-opportunity-discoverer

Amazon Opportunity Discoverer — Niche Scanner & Scoring

Tell me your budget and experience. I find opportunities, score them, and rank.

Files

• Script: {skill_base_dir}/scripts/zoodata.py — run --help for params
• Reference: {skill_base_dir}/references/reference.md (field names & response structure)

Credential

Required: ZOODATA_API_KEY. Get free key at zoodata.ai/api-keys

Capabilities & Data Flow

• Network: only https://api.zoodata.ai (Bearer ZOODATA_API_KEY). Setting ZOODATA_BASE_URL to an untrusted host (anything other than api.zoodata.ai / *.zoodata.ai / localhost) makes the CLI refuse the request and withhold the key — the Bearer token is never sent to an untrusted host.
• Execution: bundled shared ZooData CLI {skill_base_dir}/scripts/zoodata.py (Python 3, stdlib-only). This skill allows opportunity-scan, categories, market, products, product, check, plus the review fallback toolkit (reviews-raw / review-tag-prompt / review-reduce-prompt / review-aggregate). Do not invoke unrelated subcommands for this skill's tasks — the bundled manifest {skill_base_dir}/scripts/allowed-commands.json enforces this: the CLI refuses out-of-scope subcommands with a structured COMMAND_NOT_ALLOWED error before any API request.
• Local files: a private temporary working dir (created with mktemp -d, removed when the fallback completes) during the review fallback; reads the optional credential store ~/.zoodata/config.json.
• Sent to the API: keywords, category paths, ASINs, marketplace/date and numeric filter values only. Never sent: budget, experience level, risk tolerance, or any other user-profile text — profile inputs map client-side to numeric filters.
• Credits: every API call consumes account credits. For broad or ambiguous requests, state the estimated credit cost and confirm with the user before running multi-call scans. The composite opportunity-scan command executes ~15+ API calls across up to 9 loops (~25-30 credits observed) in ONE invocation and has NO skip/trim flags — under a credit cap, use the granular commands instead.

Shared CLI Contract

Before selecting or invoking the first command, read and apply the local references/cli-contract.md. Reapply it after every granular or composite result and before any fallback, additional call, state write, interpretation, or user-facing report. Use this skill's fallback logic only when the shared contract classifies the result as non-terminal.

Local Interface Failure Output

For a terminal interface failure, respond in the user's language that the opportunity scan could not be completed, followed by the succeeded and failed endpoint identifiers. Do not rank candidates, assign opportunity scores/tiers, or recommend samples or launches. Keep control tokens, parameters, and retry logs internal unless diagnostics are requested.

Input

• Required: keyword or category + budget (Low/Med/High) + experience (Beginner/Intermediate/Advanced)
• Recommended: risk tolerance (Conservative/Moderate/Aggressive)
• Optional: fulfillment preference (FBA/FBM), specific filter criteria

API Pitfalls (CRITICAL)

• categoryPath is auto-resolved via categories, with fallback to top search result. If category_source is inferred_from_search, confirm with user — keyword-only queries contaminate results
• All keyword-based endpoints MUST include --category when locked
• mode/--sales-min/--ratings-max are CLI-local, expanded client-side — NOT API fields. A raw request must use expanded API filters, must not send mode/salesMin/ratingsMax, and must distinguish ratingMax from ratingCountMax; otherwise the API returns 422.
• Revenue = sampleAvgMonthlyRevenue directly. Sales = monthlySalesFloor (lower bound)
• reviews/analysis needs 50+ reviews. Fallback chain when sample is insufficient:
  1. Lightweight: realtime/product ratingBreakdown — only star distribution, no themes
  2. Full 11-dim insights — bypass /reviews/analysis entirely: a. zoodata.py reviews-raw --asin X → fetch up to 100 raw reviews (10 credits, ~60s) b. For each review: render Map prompt via zoodata.py review-tag-prompt --review '' and have your own LLM produce JSON tags (sentiment + 11 dimensions) c. Collect candidate phrases per dimension; for each dimension render Reduce prompt via zoodata.py review-reduce-prompt --label-type X --candidates '[...]' and have your LLM produce semantic clusters d. zoodata.py review-aggregate --reviews R --tagged T --clusters C → consumerInsights output compatible with /reviews/analysis
  3. Fallback caveats (apply to the 4-step chain above — lessons from end-to-end validation):
     • Working dir: WORK=$(mktemp -d) (private, 0700 — not a predictable path); remove it with rm -rf "$WORK" after review-aggregate succeeds or the fallback aborts
     • Step b CLI behavior: review-tag-prompt RENDERS the prompt only; YOUR LLM produces the JSON. Render once to learn the schema, then produce tags for all N reviews in one in-context pass (don't call the CLI N times).
     • Step c candidate extraction (Python one-liner): candidates = {d: sorted({el.strip().lower() for t in tagged for el in (t.get(d) or [])}) for d in DIMS}
     • Small-sample rule (reviewCount<50): demote single-mention items 📊→🔍; NEVER attach table-level or section-header 📊 when any row inside is 🔍; suppress "🔴 Critical" verdicts on count=1
     • Scope: fallback replaces ONLY the /reviews/analysis aggregation. This skill's primary workflow outputs (opportunity scoring, mode-based selection, ranked candidate list) remain valid — do not re-run them.
• Deduplicate ASINs across modes — same product appears in multiple scans
• Each mode has built-in filters that STACK with user filters (e.g. high-demand-low-barrier: sales≥300, reviews≤50)

On Missing Key

When ZOODATA_API_KEY is not set (verify via python {skill_base_dir}/scripts/zoodata.py check — exits 2 if no key in env or ~/.zoodata/config.json), stop before any evidence call. Tell the user that a ZooData API key is required, link to https://zoodata.ai/en/api-keys, and explain that the key may be set in the environment or local config. Do not substitute public knowledge or a "for reference only" analysis.

On 401 Invalid Key

When _transport.status=401, stop further calls, tell the user that the configured key was rejected, direct them to https://zoodata.ai/en/api-keys, and do not fabricate missing data.

On 402 Credit Exhausted

When _transport.status=402, stop further calls. Report where the workflow stopped, any compatible partial findings already gathered, and returned credit metadata when present; direct the user to https://zoodata.ai/en/pricing and do not fabricate missing data.

Unique Logic

Profile → Strategy Mapping

User Criteria → Filter Params

Always translate: "300+ monthly sales" → --sales-min 300, "reviews <100" → --ratings-max 100, "$15-35" → --price-min 15 --price-max 35. If user has specific criteria, use custom filters (Approach B/C), NOT default modes. (--sales-min/--ratings-max/--modes are CLI-local — see API Pitfalls before any raw call.)

Data-Driven Category Selection (no specific category given)

Scan with market --keyword "{broad}" --topn 10, rank subcategories by: newSkuRate>10%, topBrandSalesRate<60%, fbaRate>50%, avgPrice $10-50, avgMonthlySales>200. Pick top 3-5.

Opportunity Score (per candidate, 1-100)

Tiers

Quick-Scan Mode (~10 credits): 2 modes × 1 page, skip realtime/trend. Label as "directional only." Implementation: run per-mode products --mode --page-size 20 calls — do NOT use the opportunity-scan composite for Quick-Scan (it always executes the full 6-step pipeline including realtime×10 + trend + reviews, ~25-30 credits, and has no skip flags).

Composite Command

python3 {skill_base_dir}/scripts/zoodata.py opportunity-scan --keyword "{kw}" --category "{path}" --modes "high-demand-low-barrier,emerging,underserved"

Or with custom filters: --sales-min 300 --ratings-max 100 --price-min 15 --price-max 35

Output

Respond in user's language.

Sections: Scan Summary → Top 10 Opportunities Table → Detailed Analysis (Top 3) → Category Heatmap → Risk Alerts → Next Steps (S: buy sample, A: deep-dive, B: watch) → Data Provenance → API Usage

If user provides COGS, calculate profit. User criteria override: ANY fail → CAUTION/AVOID.

Language (required)

Output language MUST match the user's input language. If the user asks in Chinese, the entire report is in Chinese. If in English, output in English. Exception: API field names (e.g. monthlySalesFloor, categoryPath), endpoint names, technical terms (e.g. ASIN, BSR, CR10, FBA, credits) remain in English.

Disclaimer (required, at the top of every report)

Data is based on ZooData API sampling as of [date]. Monthly sales (monthlySalesFloor) are lower-bound estimates. This analysis is for reference only and should not be the sole basis for business decisions. Validate with additional sources before acting.

Confidence Labels (required, tag EVERY conclusion)

• 📊 Data-backed — direct API data (e.g. "CR10 = 54.8% 📊")
• 🔍 Inferred — logical reasoning from data (e.g. "brand concentration is moderate 🔍")
• 💡 Directional — suggestions, predictions, strategy (e.g. "consider entering $10-15 band 💡")

Rules: Strategy recommendations are NEVER 📊. Anomalies (>200% growth) are always 💡. User criteria override AI judgment.

Aggregate-label rule (applies to ALL report output, not just fallback): NEVER attach 📊 to ANY element that aggregates or groups underlying content when ANY piece of that content is 🔍 or 💡. "Aggregate/grouping elements" include:

• Section headers at EVERY level (#, ##, ###, ####) — including top-level summary sections like "Overall Score", "Verdict", "Executive Summary"
• Summary/score lines anywhere in the report (e.g. ## Overall Score — 27/100 · Grade F 📊 is WRONG if any Basis row inside is 🔍)
• Table column headers in comparison tables (e.g. **Target ASIN** 📊 as a column label is WRONG if any cell in that column contains 🔍)
• Table row headers or row-aggregation labels (when the row aggregates multiple cells of mixed confidence)
• Any other visual grouping label — bullet-list group titles, callout box titles, etc.

A group-level 📊 implies the whole block/column/row is data-backed, which smuggles inferred/directional content into the 📊 tier via visual grouping. Either (a) omit the group-level label entirely (preferred when content mixes tiers), or (b) use the LOWEST confidence present inside (🔍 if any underlying content is 🔍; 💡 if any is 💡). This is a universal output-quality rule — it applies regardless of which fallback path (if any) was triggered.

Emoji reservation rule (closely related): The three confidence symbols 📊 🔍 💡 are RESERVED for confidence labeling. NEVER use them as decorative prefixes on section headers, table headers, or any aggregate element — even when you also include a correct confidence suffix on the same line. Example:

• ❌ WRONG: ## 📊 Overall Score — 27/100 · Grade F 🔍 (the leading 📊 reads as a data-backed claim even though the trailing 🔍 is correct)
• ✅ RIGHT: ## Overall Score — 27/100 · Grade F 🔍 (no decorative emoji, just the proper confidence suffix)
• ✅ RIGHT: ## 🎯 Overall Score — 27/100 · Grade F 🔍 (use non-reserved decorative icons like 🎯 🧭 📋 📝 📂 🏁 🚨 🏆 🔔 when a visual prefix is desired)

Decorative emoji ≠ confidence label — but from a reader's perspective, a leading 📊/🔍/💡 is indistinguishable from a confidence claim. Reserve these three symbols EXCLUSIVELY for confidence annotation to avoid ambiguity.

Data Provenance (required)

Include a table at the end of every report:

Extract endpoint and params from _query in JSON output. Add notes: sampling method, T+1 delay, realtime vs DB, minimum review threshold, etc.

API Usage (required)

Extract from meta.creditsConsumed per response. End with Credits remaining: N.

API Budget: ~50-60 credits