agentleFS
proyecto26proyecto26/instagram-ai-plugin/skills/instagram-ads/SKILL.md

instagram-ads

Skill1 starsChanged 12 days agoMITView on GitHub

Plans and evaluates paid campaigns that run on Instagram (Meta Ads with Instagram feed, Stories and Reels placements). It picks which organic posts to promote using real performance pulled through the user's web session, writes ad copy and creative briefs, sets up Andromeda-era campaign structure, and makes kill, keep and scale decisions from numbers the user pastes. This skill should be used when the user asks to "run Instagram ads", "boost this post", "which post should I promote", "write Instagram ad copy", "set up a Meta campaign for Instagram", "is my Instagram ad working", "when should I kill this ad", "creative testing for Reels ads", or mentions ROAS, CPA, CPM or partnership ads on Instagram.

The file

We split this skill into the rules it gives an agent

8 rules. About 0 are common, 4 are distinctive, and 4 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈4

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 4

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Use organic evidence to choose what to pay for, structure the campaign so Meta's delivery system can learn, and decide with numbers instead …

    line 8
  2. Very specific:

    Can: read the account's organic performance (instagram skill audit, posts, media, and swipe for competitors), write copy and briefs, and …

    line 13
  3. Distinctive:

    Cannot: create, edit, boost, pause or fund ads.

    line 15
  4. Distinctive:

    Never estimate missing ad metrics.

    line 19
  5. Very specific:

    Read .agents/product-marketing.md (or .claude/product-marketing.md) if present.

    line 23
  6. Very specific:

    When the account has no outliers, build concepts from a competitor swipe file (instagram-research) and adapt their structure, never their …

    line 40
  7. Distinctive:

    Do not retarget with the same ad.

    line 56
  8. Very specific:

    Ask for these per ad, last 7 to 14 days: spend, impressions, CPM, link CTR, CPC, conversions, CPA or ROAS, frequency, and 3-second hold for …

    line 69

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 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.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Instagram ads

Use organic evidence to choose what to pay for, structure the campaign so Meta's delivery system can learn, and decide with numbers instead of feel.

What this plugin can and cannot do

  • Can: read the account's organic performance (instagram skill audit, posts, media, and swipe for competitors), write copy and briefs, and evaluate exported or pasted results.
  • Cannot: create, edit, boost, pause or fund ads. Ads Manager and the Boost flow use Meta's business systems and payment methods. The Instagram web cookie does not drive them, and this plugin must never try. The user makes every change in Ads Manager or the app, so give exact settings to enter.
  • Never estimate missing ad metrics. Label anything not provided MISSING and say what it would change.

Before starting

Read .agents/product-marketing.md (or .claude/product-marketing.md) if present. Collect only what is missing: the goal (sales, leads, followers, traffic), the offer and price, the target CPA or ROAS, the monthly budget, the pixel or Conversions API status, and any current results.

1. Choose what to promote (organic first)

From the instagram skill directory:

npx -y bun scripts/main.ts audit --limit 30 --json > "<project-dir>/instagram-audit.json"   # user folder, not the plugin

Shortlist outliers (3x or more of the account median). A post that already beats the account's own median organically is the cheapest creative test there is. Prefer posts whose hook works with the sound off, and whose first 3 seconds or cover state the claim. Comments full of questions are a signal of buying intent. Read them with comments <url>.

When the account has no outliers, build concepts from a competitor swipe file (instagram-research) and adapt their structure, never their copy.

2. Structure (Andromeda era)

Summary. The full detail is in references/meta-andromeda-playbook.md and references/meta-decision-system.md.

  • Broad audience, specific creative. The creative does the targeting. Avoid stacked interests.
  • Creative volume is the lever. Ship many cheap, distinct concepts (statics plus native-looking Reels) rather than polishing one. Variants of one idea count as one concept.
  • Two campaigns: Testing (new concepts, controlled budget) and Scaling (graduated winners).
  • Placements: Instagram Reels and Stories need 9:16 with text inside the safe zone (../instagram-content/references/formats.md). Feed takes 4:5.
  • Partnership ads (creator posts promoted through the creator's handle) add net-new reach. They need the creator's permission in the app.
  • Sequence: cold concept, then a different-angle retarget (proof or objection), then the offer. Do not retarget with the same ad.

3. Write the ads

For each concept give: the hook (first line or first 3 s), primary text (lead with the outcome, at most 125 chars before "more"), the headline, the CTA button, the creative brief (format, first frame, on-screen text, length) and the placement. Templates are in references/ad-copy-templates.md. Audience notes are in references/audience-targeting.md. Lint the primary text with npx -y bun ../instagram-content/scripts/caption-lint.ts primary.txt and npx -y bun ../instagram-content/scripts/ai-tells.ts detect primary.txt. Native beats polished: ads that look like ads get scrolled past.

4. Decide: kill, keep, scale

Ask for these per ad, last 7 to 14 days: spend, impressions, CPM, link CTR, CPC, conversions, CPA or ROAS, frequency, and 3-second hold for video. Then apply references/meta-decision-system.md:

  • Day 7 delivery check: an ad with almost no spend is a delivery verdict, not a creative one.
  • Weekly quality check on rolling 14-day data against the true CPA or CPL target.
  • Fatigue: frequency rising, CTR falling, and CPA climbing together mean refresh the concept.
  • Graduate to Scaling only past the documented thresholds. Scale budgets in steps, not jumps.

5. Output

FACTS        what the numbers show (with the date range and source)
HYPOTHESES   what probably explains it, with confidence
MISSING      inputs not provided and what they would change
ACTIONS      exact changes for the user to make in Ads Manager, in order

End every recommendation with those four blocks.

Additional resources

  • references/meta-andromeda-playbook.md: creative volume, creative as targeting, creator ads, sequencing.
  • references/meta-decision-system.md: TCPL, two-campaign structure, delivery and quality stages, fatigue, scaling, partnership ads, benchmarks.
  • references/ad-copy-templates.md: copy formulas and templates.
  • references/audience-targeting.md: audience research and targeting by platform.

Credits: the references are copied from marketingskills ads (MIT, Corey Haines); see the plugin's THIRD_PARTY_NOTICES.md.

Download now (ZIP)

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
What did you use it for? e.g. Reviewing pull requests in a Rails app

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.