record-creative-learning
Skill0 starsChanged 27 days agoMITView on GitHub
Close the creative loop: turn your own ad results (an ads manager export, a report, numbers you paste) or an audit into a scoped learning (the why, not the what), write it back into the next concepts, and find AdSwipeFile reference ads for what to make next. Use after a creative test ends, when asked "what did we learn", "why did this ad do well", or before a new round of concepts.
In this repository14 other files it gives its agents.
- audit-ad-creativeSkill
- competitor-adsSkill
- creative-strategySkill
- find-winning-adsSkill
- generate-ad-conceptsSkill
- get-startedSkill
- map-competitorsSkill
- map-personasSkill
- pick-creative-mechanicSkill
- review-saved-adsSkill
- teach-ad-breakdownSkill
- track-competitor-adsSkill
- weekly-competitor-reviewSkill
- write-hooks-from-adsSkill
The file
We split this skill into the rules it gives an agent
5 rules. About 0 are common, 3 are distinctive, and 2 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈3
Distinctive · estimate
This author's own choices. Worth a look.
2
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 28
Ask what each creative was (persona, angle, format, hook) when the export doesn't say.
- Very specific: line 70
Read three to five with get_ads for their breakdown and why_it_works.
- Very specific: line 72
When the user asks, file them on a board (create_board, then add_ad_to_board) and put the learning's implication on each as set_ad_note …
- Distinctive: line 77
Never borrow a library ad's results: it has none.
- Distinctive: line 79
Never invent the user's numbers.
≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Record a creative learning
Needs the AdSwipeFile connector. This skill reads the AdSwipeFile ad library of Meta ads through its MCP server. If no AdSwipeFile tools (
search_library,list_ads) are available, say so before anything else and give the step for this host. Claude Code:claude mcp add --transport http adswipefile https://adswipefile.com/api/mcp. claude.ai or ChatGPT: addhttps://adswipefile.com/api/mcpas a custom connector. Other clients: https://adswipefile.com/docs/mcp?utm_source=agent-skill&utm_medium=referral&utm_campaign=record-creative-learning. Connecting opens a sign-in to an AdSwipeFile account. Steps that need only the user's own material can go on; never invent ads, brands or links in place of the library.
Research feeds concepts, concepts become tests, and what tests show feeds the next research. If
nothing in the personas, angles or briefs ever changes, nobody is looking. Words such as persona,
angle, anchor and mechanic are used as the creative-strategy skill's glossary defines them.
AdSwipeFile holds no results, for the user's ads or anyone's. The results come from the user: an ads manager export, a report, numbers they paste, or another connector they have. Ask for them when they're missing.
1. Gather what happened
- The results: which creatives ran, against what, for how long, on which goal, and the numbers. Ask what each creative was (persona, angle, format, hook) when the export doesn't say.
- Read it honestly. One creative is an anecdote. A pattern across several comparable creatives (same goal, similar audience and spend) is worth a hypothesis. Low delivery is inconclusive, never a loss. An inconclusive test teaches about the test (too little spend, too many variables at once), not about the creative.
- Earlier learnings the user keeps: a new finding on the same scope updates that learning, it doesn't add a second.
2. Write the why
Each learning has four parts:
- Statement: what changed in the viewer's mind, brand-specific, still true with the numbers removed ("night-shift workers trust a coworker's recommendation over a doctor's"). If it only restates a number, rewrite it.
- Observation: what happened, in the numbers' terms, with where the numbers came from.
- Implication: what to make differently. Briefs quote this line.
- Scope: what it applies to: persona, angle, pain, awareness stage, format, hook, mechanic. AND across them, OR within one ("persona: night-shift nurses; format: testimonial or street interview").
Mark it a hypothesis until a clean test backs it.
3. Write it back
Each learning lands somewhere a later concept will read it:
- An angle that keeps landing: name the personas and stages it hasn't been made for yet, as the next
concepts (
generate-ad-concepts). - A trigger moment that keeps showing up: add it to that persona's moments.
- A format or mechanic that keeps falling short at one stage: say so in the implication, so the next concepts skip it there.
- A hook line customers repeat back: keep it as a hook on its angle (
write-hooks-from-ads). - A refuted learning stays, marked refuted, so the next test that asks the same question sees it.
4. Make it actionable on AdSwipeFile
Turn the implication into references to make the next creatives from:
- Match its scope to library tags:
list_visual_formatsandlist_hooksfor the slugs. list_adsbyvisual_formatorhook(andnichewhen it fits),sort: "longest_running",limitabout 10.- Read three to five with
get_adsfor their breakdown andwhy_it_works. Pick the ones that show the move the implication asks for, and link each by itsurl. - When the user asks, file them on a board (
create_board, thenadd_ad_to_board) and put the learning's implication on each asset_ad_notewithboard.
Rules
- Never borrow a library ad's results: it has none. A long run is a reason to study it; write "kept running for 94 days".
- Never invent the user's numbers. Quote them as given, with their source.
- Return the learnings in the reply (or a file, if the client can write one).
Done when each finding is one learning on the right scope, its implication says what to make next, and reference ads for that next step are linked or filed where the user asked.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

