Grow the account: 10 AI skills for social media

profile-optimiser

a bio and header that convert visitors into followers

How the two work together

Claude thinks it through. Paste the Claude prompt into Claude Code, or drop the folder into your skills folder. Claude does the judgement: what to look for, what is worth doing, what is right.

Codex gets it done. At the hand-off point Claude runs Codex on your machine with one command and passes it the Codex prompt. Codex does the mechanical part and hands the result back. Claude checks it before you see it.

No API key to set up: Claude calls the Codex you already have installed. If Codex is not installed, Claude does that half itself and tells you.

Prompt for Claude

---
name: profile-optimiser
description: Rewrites a LinkedIn or Instagram headline, bio and link to real character limits and 2026 engagement data. Use when a profile gets views but few follows.
---

# A bio and header that convert visitors into followers

A headline, bio and link that turn profile views into follows.

## Claude does

1. Ask for four things and refuse to start without them: the profile URL, the one outcome a visitor gets (not what you sell), three evidenced proof facts, and the single action a visitor should take.
2. Read the live profile. Copy out verbatim the current headline, bio, link and first three posts. Never work from memory.
3. Pick the surface. Personal profiles beat company pages by 63% on engagement (Metricool, 673,658 posts), so optimise the personal one first and say so.
4. Write to hard counts, not adjectives. Headline: 8 to 12 words, outcome inside the first 50 characters, the mobile search cut. Instagram bio: 150 characters, outcome inside the first 80. LinkedIn About: outcome and one number above the two-line fold.
5. One formula, varied inputs: who you help + the outcome + a proof number. Never open with a job title or a tool name.
6. Score drafts against the account's own top 10% of posts, ranked by reactions + (comments x 3), the method in charlie947/social-media-skills. The profile should sound like what already works.
7. Choose one link, named by what it gives. Refuse to invent a URL, a follower count, a client name or a result the user did not supply.

## Then Codex does

Codex takes the mechanical bulk: the variant grid, character counts enforced programmatically, truncation previews at each cut point, the before/after table. Counting at volume is where a chat drifts.

```
codex exec --sandbox danger-full-access --skip-git-repo-check -C "<working folder>" - < CODEX.md
```

Fill in first: {PLATFORM}, {CURRENT_PROFILE_TEXT}, {AUDIENCE}, {OUTCOME}, {PROOF_FACTS}, {ONE_ACTION}, {LINK_URL}.

## Claude checks

1. Count the characters yourself. Never trust a stated count.
2. Read each truncation preview: does the outcome survive the cut, or does the fold land mid-sentence?
3. Check every number and name against {PROOF_FACTS}. Anything else is invented, so delete it. Caption repos circulate precise-sounding statistics with no source; this is how they get in.
4. Confirm no hashtags were added. Hashtags cost 31.70% of views (Metricool, 24,364,803 posts).
5. Confirm one link, live, named by its benefit.
6. Any failure: rerun Codex quoting the failing item. Never hand-patch the output and present it as generated.

## Rules

- Personal profile before company page.
- Hard counts, never adjectives. "Punchy" is not a specification.
- No hashtags in a bio or headline.
- Never cite a statistic you cannot trace to a named study with a sample size.
- Public information only.
- Never invent a fact, a number or a quote.
- Anything sent in someone's name says whose name it is.

## If Codex is not installed

Claude builds the variant grid and does the counting itself, and says so.

## Built from

- charlie947/social-media-skills, https://github.com/charlie947/social-media-skills, 3,283 stars via api.github.com; profile-optimizer and post-scorer, behind a 415k+ follower system.
- Metricool 2026 LinkedIn Study, https://metricool.com/press-release-linkedin-study-2026/, 673,658 posts; personal profiles +63%.
- Metricool 2026 Instagram Study, https://metricool.com/press-release-instagram-study-2026/, 24,364,803 posts; hashtags -31.70% views.
- coreyhaines31/marketingskills, https://github.com/coreyhaines31/marketingskills, 46,945 stars via api.github.com; the persistent context-file pattern.

Prompt for Codex

# A bio and header that convert visitors into followers - Codex task

## You are given

- {PLATFORM}: linkedin, instagram, or both.
- {CURRENT_PROFILE_TEXT}: the existing headline, bio and featured link, copied verbatim from the live profile.
- {AUDIENCE}: who the profile is for, in the user's own words.
- {OUTCOME}: the single result a visitor gets by following. Not a job title, not a tool name.
- {PROOF_FACTS}: three evidenced facts (numbers, names, dates). This is the ONLY pool of facts you may use.
- {ONE_ACTION}: the single action a new visitor should take.
- {LINK_URL}: one destination URL.

## Produce

Write `profile-options.md` in the working folder containing, in this order:

1. **Before table.** {CURRENT_PROFILE_TEXT} split into headline / bio / link, each with its exact character count.
2. **5 headline variants.** Each 8 to 12 words. Each with the first 50 characters marked with a `|` at the cut point, then the full count in brackets. Formula for every one: who you help + the outcome + a proof number from {PROOF_FACTS}.
3. **3 bio variants per platform in {PLATFORM}.**
   - Instagram: exactly 150 characters or fewer, {OUTCOME} inside the first 80. Mark character 80 with `|`.
   - LinkedIn About: 3 short paragraphs. The first 2 lines must carry {OUTCOME} and one proof number. Mark the 2-line fold with `--- fold ---`.
4. **Truncation preview block.** For every variant, print exactly the text a visitor sees before the platform cuts it: 50 chars for a LinkedIn headline in search, 80 chars for an Instagram bio line, 2 lines for a LinkedIn About.
5. **Link line.** One line naming {LINK_URL} by what it gives the visitor, under 40 characters, no "link in bio".
6. **Counts table.** Every variant, its character count, and PASS or FAIL against its limit.

## Rules

- Every number, name, client and date must come from {PROOF_FACTS}. Invent nothing.
- No hashtags anywhere in a bio or headline.
- No emoji unless one already appears in {CURRENT_PROFILE_TEXT}.
- British English.
- No em dash characters.
- Count characters programmatically, not by eye. Include spaces and punctuation.
- One link only. Do not add a second URL, a phone number or an email.
- Do not soften or reword a proof fact to make it fit. Drop it instead.

## Return

Print to stdout:

1. The absolute path of `profile-options.md`.
2. The counts table in full.
3. A line `FACTS USED:` listing each proof fact and the variant numbers it appears in.
4. A line `LIMIT FAILURES: <n>` with each failing variant named, or `LIMIT FAILURES: 0`.

Built from the best public work on this

Sources

  • **charlie947/social-media-skills** - https://github.com/charlie947/social-media-skills - 3,283 stars and 771 forks confirmed via api.github.com and an independent raw curl parse. The working skill set behind Charlie Hills' own content system (415k+ followers, 100m+ views a year; ranked #1 worldwide for LinkedIn growth by Favikon). Supplies the `profile-optimizer` skill and the `post-scorer` scoring method used here.
  • **Metricool 2026 LinkedIn Study** - https://metricool.com/press-release-linkedin-study-2026/ - 673,658 posts from 63,108 accounts worldwide, Jan-Feb 2025 vs Jan-Feb 2026. Source of the finding that personal profiles generate 63% higher engagement than company pages.
  • **Metricool 2026 Instagram Study** - https://metricool.com/press-release-instagram-study-2026/ - 24,364,803 posts from 375,118 accounts worldwide. Source of the hashtag penalty: -31.70% views, -33.89% interactions.
  • **coreyhaines31/marketingskills** - https://github.com/coreyhaines31/marketingskills - 46,945 stars and 7,317 forks confirmed via api.github.com and raw curl parse, actively maintained. Source of the persistent context-file pattern that every skill reads before drafting.

Best public prompt we found for this job

https://github.com/charlie947/social-media-skills

The repo's architecture is the part worth copying: a persistent context layer (`about-me.md` plus `voice.md`) that every other skill reads before drafting, asking the user only for what is missing. Its `post-scorer` skill computes engagement as `reactions + (comments x 3)`, extracts the top and bottom 10% of the user's own posts, and scores new drafts against those patterns rather than against generic best practice. That is the single highest-leverage technique in the whole pack, and it applies to a profile as much as to a post: write the headline in the language that already earns this account its best engagement.

Want this running in your business?

I optimise how businesses run — your sales, your visibility, your social media — and build bespoke software where nothing off the shelf fits. The first conversation is free. Work starts from £150 a day.