Creative Concept Finder

Find the creative concepts already working in your category, and turn them into a library that accumulates.

flowchart LR
  A["<b>1 · Your brand,<br/>category, audience</b>"]
  B["<b>2 · Research, filter,<br/>dedupe</b>"]
  C["<b>3 · Briefable cards</b><br/>that record what<br/>happened"]
  A --> B --> C
  style A fill:#f6f8fa,stroke:#57606a,color:#1f2328
  style B fill:#ddf4ff,stroke:#0969da,color:#1f2328
  style C fill:#dafbe1,stroke:#1a7f37,color:#1f2328

Loading

Ask any research tool to find good ads in your category and you get a competitor summary: who is running what, roughly. None of it is usable, because a competitor's ad is an execution. What a creative team needs is the concept underneath it — the transferable idea, separated from the brand that expressed it.

Research output is a list, and a list gets read once. Cards are durable.

---

Run it

1. Pick the skill you need from the table below.
2. Put its SKILL.md in your AI tool's instructions field — Claude Project, Custom GPT, Gemini Gem, or an API system prompt.
3. Put anything in its reference/ folder, and anything in this repo's reference/, into the knowledge base.

Missing a required input, they ask instead of guessing.

---

What's in here

Skill	Does
creative-concept-researcher	Writes the research prompt, sweeps four sources, filters what comes back
creative-concept-card-builder	Turns concepts into briefable cards and a library that grows

They chain, but each one runs on its own.

---

How it works

1. The researcher states the concept-versus-execution distinction inside the prompt, because tools default to returning executions
2. It sweeps adjacent categories deliberately — an in-category search returns what your competitors already do
3. Repetition across unrelated brands is treated as evidence, so deduplication keeps the observation count
4. The card builder checks the library first and merges rather than duplicating

---

Notes

Failures are kept, in full. A concept tried and abandoned is a more valuable card than an untried one — deleting it guarantees someone proposes it again in eighteen months with the same enthusiasm and no memory.

---

Files

skills/            one folder per skill — SKILL.md plus its reference files
reference/         shared reference the skills load alongside them
LICENSE            MIT

---

MIT © Raneq Barber