Marketing Foundation
Compact two-skill starter: clarify positioning and choose a lead magnet. Use Marketing Launch for the broader eight-skill go-to-market workflow.
The tunnel uses the same cards as the catalogue. Browse only as deep as needed — or load a broad bundle immediately.
SEO, Sales, Agents or another broad area → one bundle call → work.
Read-only access to published skills. Default 8, maximum 10 skills / 120,000 characters.
Compact two-skill starter: clarify positioning and choose a lead magnet. Use Marketing Launch for the broader eight-skill go-to-market workflow.
Build an evidence-led marketing plan from ICP and competition through positioning, campaigns, growth and measurement.
Diagnose architecture and context, plan agent-team responsibilities, then organize project context and session handoffs. Memory and cost-runtime reviews remain outside this pack.
Review the journey from landing page and lead capture through registration, first value and transparent upgrades.
Plan a campaign, draft its channel content and review the work against actual brand guidance.
Prioritize an editorial roadmap and plan how to launch and distribute it across suitable channels.
Understand customer needs, compare competitors and plan a community around real member value.
Choose a relevant lead magnet, then draft a permission-based nurture journey with entry, suppression and exit rules.
Define the API contract, then plan how to observe its latency, failures and retries. Guidance and checklist; no production changes.
Profile a dataset, choose and interpret statistical methods, then validate calculations and conclusions before sharing.
Combine competitor, own-creative and hook evidence into test directions; this synthesizes existing research rather than collecting new data.
--- name: winning-pattern-synthesis description: > Cross-reference competitor ad patterns against the user's own strong-performing creatives and psychological hook analysis to find cross-source overlap, own-account test gaps, and market whitespace. Use whenever the user wants to know what to actually build next — after competitor research and/or own-creative diagnosis have already run, or when they explicitly ask "what should we test" or "what's the whitespace." This is a synthesis skill — it consumes the outputs of competitor-ad-intelligence, own-creative-diagnosis, and creative-psychology-hooks rather than pulling new data itself. --- # Winning Pattern Synthesis Takes the outputs of the other three analysis skills and turns them into a ranked set of creative directions worth actually building — the step between "here's what we found" and "here's what to brief." ## When to use this skill - "What should we test next?" - "Where's the whitespace vs competitors?" - Any time competitor intel + own-creative diagnosis have both run and the user wants a single synthesized direction rather than two separate reports - As the step before `ad-brief-generator` ## Inputs This skill doesn't fetch new data — it reasons over what's already been produced: - **Competitor patterns** from `competitor-ad-intelligence` (angles, hooks, formats, scale signals) - **Own performance patterns** from `own-creative-diagnosis` (what showed strong first-party results in the stated analysis window) - **Psychological grading** from `creative-psychology-hooks` (mechanism, valence/intensity, micro-moment) If one or more inputs are missing, say explicitly which are missing and whether to proceed with a partial synthesis or run the missing skill(s) first. Competitor-only or own-creative-only synthesis is possible but weaker — flag it as partial. ## Step 1: Build the pattern matrix Lay out every distinct pattern (angle × hook mechanism × format) found across both sources in one table, tagging each occurrence: | Pattern | Seen in competitors? | Seen in own strong performers? | Scale/strength signal | |---|---|---|---| ## Step 2: Classify each pattern - **Cross-source overlap** — the pattern has strong first-party results and competitor duration/spend/variation proxy evidence. This is the strongest observed overlap, but competitor performance remains unverified; recommend a controlled variation test rather than calling it proven. - **Competitor-backed test gap** — competitors are running it with meaningful proxy signals, but the user hasn't tested it. This is an own-account test gap, not market whitespace or a guaranteed winner. - **Own-only market whitespace** — has strong first-party results in the analysis window but was not observed in the competitor sample. It could be a differentiator, a sample-coverage gap, or an account-specific result — note all applicable interpretations. - **Neither** — untested by anyone visible. Highest risk/highest potential differentiation; only surface if the user wants aggressive experimentation, not as a default recommendation. ## Step 3: Rank by confidence and apply psychological lens For each pattern surfaced, pull in the Baader mechanism / valence-intensity classification from `creative-psychology-hooks` if hooks for that pattern were graded. Present it as a testable hypothesis for why the pattern may attract attention, not a causal explanation of performance. Rank output: cross-source overlap first, then competitor-backed test gaps, then own-only market whitespace. Cap at the top 5–7 directions unless the user asks for the full list — a synthesis that recommends everything recommends nothing. ## Step 4: Flag conflicts If a pattern is a competitor favorite but has performed poorly in the user's own account (or vice versa), don't average it out — call out the conflict directly and suggest a hypothesis for the divergence (audience difference, execution quality, offer mismatch). ## Output Format ``` ### Winning Pattern Synthesis: [Product/Category] Inputs used: [competitor intel / own diagnosis / psych grading — note any missing] Top directions: 1. [Pattern] — CROSS-SOURCE OVERLAP - Competitor signal: [brands, scale] - Own signal: [ad, metrics] - Mechanism: [Baader tag, valence/intensity] - Recommendation: [controlled variation test / new variant angle] 2. [Pattern] — COMPETITOR-BACKED TEST GAP - ... Conflicts worth noting: - [Pattern]: strong for competitors, weak in our account — possible reason: [hypothesis] Not recommended right now: - [pattern seen but low-confidence, and why] ``` ## Guardrails - Never present competitor proxy evidence as verified performance or competitor-backed patterns as proven for the user's account. - Don't let this skill turn into a restatement of the two input reports — its entire value is the cross-reference and ranking. If there's no real overlap or conflict to surface, say that plainly rather than padding the output. - Hand off the final ranked list directly to `ad-brief-generator` — don't re-derive angles from scratch there.
Combine competitor, own-creative and hook evidence into test directions; this synthesizes existing research rather than collecting new data.
The complete original GoMarble guidance, with source attribution and declared limits.
Clarify scope, use available evidence, and apply the relevant source workflow within actual user authorization.
Competitor research, own-creative findings, hook analysis and actual test constraints.
Overlap across samples does not establish causation or market-wide whitespace; label absent and contradictory evidence. GoMarble connectors, referenced sibling skills and runtime dependencies are not bundled or installed by M11. Brief obvious-danger screening only; no functional test or comprehensive safety certification. Loading this text authorizes no external calls, account mutations or ad spend.
Use Ad Pattern Synthesis for [TASK]. Clarify Competitor research, own-creative findings, hook analysis and actual test constraints. Separate observed evidence from heuristics and recommendations. Check actual connector availability and approval semantics before any external action.
Claiming installed GoMarble tools, proven campaign outcomes or authorization to launch ads merely from loading this source.
German routing and explicit scope. Original attribution: GoMarble. Overlap across samples does not establish causation or market-wide whitespace; label absent and contradictory evidence. GoMarble connectors, referenced sibling skills and runtime dependencies are not bundled or installed by M11. Brief obvious-danger screening only; no functional test or comprehensive safety certification. Loading this text authorizes no external calls, account mutations or ad spend.
MIT License Copyright (c) 2026 GoMarble Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions: The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software. THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
Copy the text below, then paste it into your chat.