17 AI skills for finding customers and making content
comment-to-dm-post
get comments and DMs from a post - the format that brought you here
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: comment-to-dm-post
description: Write a comment-a-word-and-I'll-send-it giveaway post, the DM that delivers it and the replies that keep it alive; use when giving away a real template, checklist, prompt pack or playbook to start DM conversations and grow followers.
---
# Get comments and DMs from one giveaway post
One post, a thread full of your keyword, and a DM conversation with everyone who asked.
## Claude does
1. Get from the user: pack name, link, every piece inside, who it is for, the one problem it solves, one true number about their experience, when they last ran this format, and a screenshot of the pack for the image.
2. Open the link and check every piece exists and works. If not, stop and say so.
3. Qualify the pack: one specific problem, usable in under 30 minutes, one format, phone-readable, a natural step to the paid offer. Prompt packs, playbooks, checklists and templates earn comments; ebooks, courses and webinar replays do not. Name the failed test and stop. Never invent the number.
4. Refuse the format if they ran it within the last month, or the audience is developers, CTOs or other experts (trust cost beats reach). Offer a plain value post instead.
5. Turn each piece into a one-line outcome ("get cold emails answered", not "email_template_v2"). 3 to 5 bullets; a named set may list up to 12; beyond that, group.
6. Pick {WORD}: one word, capitals, easy to spell, tied to the topic, unused on earlier posts. Set {SCARCITY} (a light line or "none") and {QUESTION}, one genuine question for the DM.
7. Fix the ten values below.
## Then Codex does
Codex drafts to a fixed skeleton with hard character limits: post, delivery DM, ten public replies, day-three follow-up, tracker. Volume drafting to a decided shape is Codex work.
```
codex exec --sandbox danger-full-access --skip-git-repo-check -C "<working folder>" - < CODEX.md
```
Fill the "You are given" block first: {PACK_NAME}, {PACK_LINK}, {PROBLEM}, {AUTHORITY_NUMBER}, {PATTERN}, {ITEMS}, {HOW_IT_WORKS}, {WORD}, {SCARCITY}, {QUESTION}.
## Claude checks
1. post.md order: authority number, pattern, named bundle, "What's inside", "Free. No email wall." plus scarcity, ask. Authority number inside the first 210 characters; 900 to 1,300 characters total; double line breaks; 0 to 2 hashtags at the end.
2. No URL, one ask, no "like if you agree", "DM me to learn more" or "repost if useful".
3. {WORD} twice, quoted, capitals, exactly as given, and the connect line present (non-connections cannot receive the DM).
4. Every bullet matches a piece checked in step 2; no number or claim the user did not give; no over-selling ("the most comprehensive guide you will read this year").
5. dm-reply.md names the topic, has the exact {PACK_LINK} on its own line, asks {QUESTION}, no pitch, no booking link, under 40 words.
6. Ten public replies without links; follow-up keeps [N] and [Y]; tracker.csv has six columns.
Any failure: list it at the top of CODEX.md under "Fix these", rerun, recheck. Nothing reaches the user until all six pass.
## Rules
- The pack must be real and must work. A weak pack costs more followers than it gains.
- Never a bare "comment X" line; the gate sits inside a substantive post or LinkedIn suppresses it.
- One ask, one keyword, one word. No link in the body (reach cut 40 to 60 per cent). No landing page or email form before delivery.
- The connect line goes in every time.
- DM within an hour of the comment. Reply to every comment within 90 minutes; a one-line public reply keeps the post alive.
- The DM hands over the pack and asks one question. Never a pitch or a booking link in its place.
- At most once a month.
- Track comments, DMs sent, connections gained, replies to the DM, calls booked. Comments are not demand; replies and calls are.
- Post the follow-up three days later with the real [N] and the item most people used first.
## If Codex is not installed
Claude does the "Then Codex does" section itself, says so, and still runs "Claude checks".
## Built from
- sergebulaev/linkedin-skills, F6 "Comment-Gate Lead Magnet": https://github.com/sergebulaev/linkedin-skills/blob/main/references/hook-formulas.md (868 stars on api.github.com, 5 Sep 2026; mirrored with credit in sickn33/agentic-awesome-skills, 45,983 stars).
- coreyhaines31/marketingskills, lead-magnets: https://github.com/coreyhaines31/marketingskills/blob/main/skills/lead-magnets/SKILL.md (46,908 stars on api.github.com, 5 Sep 2026).
Prompt for Codex
# Get comments and DMs from one giveaway post - Codex task
You are drafting a LinkedIn giveaway post: the reader comments a keyword, the sender delivers the pack by DM. Claude has checked the pack and fixed every value below. Use only what is given.
## You are given
- {PACK_NAME}: the named bundle
- {PACK_LINK}: the exact URL that delivers it
- {PROBLEM}: the one problem it solves, in the reader's words
- {AUTHORITY_NUMBER}: one true sentence containing a real number
- {PATTERN}: the one-sentence observation that number earned
- {ITEMS}: 3 to 12 one-line outcomes, one per line, in order, final
- {HOW_IT_WORKS}: one line on what the reader does with the pack
- {WORD}: the comment keyword, one word, capitals
- {SCARCITY}: a light scarcity line, or "none"
- {QUESTION}: the one question the DM asks
## Produce
Five files in the working folder.
1. `post.md`, in this order, one blank line between parts, nothing else:
1. {AUTHORITY_NUMBER}
2. {PATTERN}
3. So I turned that into {PACK_NAME}: N pieces for {PROBLEM}. Comment "{WORD}" and it's yours. (N = count of {ITEMS}.) Then {HOW_IT_WORKS} on the next line.
4. "What's inside:" then {ITEMS}, one per line starting "- ", exact wording and order.
5. "Free. No email wall." plus {SCARCITY} on the same line unless "none".
6. Comment "{WORD}" below + connect with me and I'll send it.
7. Optional last line: up to two hashtags.
Limits: {AUTHORITY_NUMBER} ends inside the first 210 characters; 900 to 1,300 characters total; {WORD} exactly twice, both in double quotes; no URL anywhere.
2. `dm-reply.md`, under 40 words: Hi {FIRST_NAME}, thanks for commenting on my post about {PROBLEM}. Here's {PACK_NAME}: then {PACK_LINK} on its own line, then {QUESTION}. Nothing after the question.
3. `public-replies.md`: ten different one-line replies, each under 12 words, each using {FIRST_NAME}, no links. Five confirm delivery ("Sent, {FIRST_NAME}. Check your DMs."); five ask the commenter to accept the connection so the DM can go through.
4. `follow-up-post.md`: opens "[N] people asked for {PACK_NAME}. The piece most of them used first was [Y]. Here's why:" then three short lines about [Y]. No ask. Keep [N] and [Y] as placeholders.
5. `tracker.csv`: header `date,comments,dms_sent,connections_gained,replies_to_dm,calls_booked` and one empty row.
## Rules
- Use only what is given. Invent no items, numbers, claims or testimonials.
- Items are outcomes. If one reads like a file name, keep it as given and flag it.
- One ask. No second call to action, newsletter line, booking link or "repost if useful".
- Never "like if you agree", "DM me to learn more" or "thoughts?".
- No superlatives about the pack ("most comprehensive", "everything you need").
- Count characters and words yourself.
## Return
Print only:
- Full path of each file written
- post.md: total characters, position where {AUTHORITY_NUMBER} ends, {WORD} count, hashtag count
- dm-reply.md: word count
- Anything that looked wrong: an item like a file name, a multi-word WORD, a link that is not a URL, an {AUTHORITY_NUMBER} with no digit
Do not print file contents. Claude reads the files.
Built from the best public work on this
Sources for comment-to-dm-post
Every source below was opened live on 5 Sep 2026. Star counts were read from api.github.com, not from the repo page.
Built from
- sergebulaev/linkedin-skills, formula F6 "Comment-Gate Lead Magnet": https://github.com/sergebulaev/linkedin-skills/blob/main/references/hook-formulas.md (868 stars, 138 forks, MIT, pushed 3 Sep 2026, per api.github.com; mirrored with credit in sickn33/agentic-awesome-skills, 45,983 stars). Gives the post skeleton, the 210/140-character hook cutoff, the 900 to 1,300 character body, double line breaks, 0 to 2 hashtags, no links in the body, reply within 90 minutes, and the once-a-month limit.
- coreyhaines31/marketingskills, lead-magnets skill: https://github.com/coreyhaines31/marketingskills/blob/main/skills/lead-magnets/SKILL.md (46,908 stars, 7,312 forks, MIT, pushed 3 Sep 2026, per api.github.com). Gives the rules for the thing being given away: one specific problem, usable in under 30 minutes, one format, mobile-friendly, a natural path to the paid product, 3 to 5 "what's inside" bullets, show a preview, ask for the minimum.
Corroborating reads (vendor articles, used to confirm the two sources above, not credited as sources)
- LinkMagnet, "How to structure a LinkedIn comment-to-DM post": https://linkmagnet.gg/blog/structurer-post-linkedin-comment-to-dm
- MagicPost, "LinkedIn Lead Magnet Posts: Examples + Templates" (11,530-post dataset): https://magicpost.in/blog/linkedin-lead-magnet
- LigoSocial, "LinkedIn Call to Action Examples: 30+ CTAs by Goal": https://ligosocial.com/blog/linkedin-call-to-action-examples
- ConnectSafely, "LinkedIn Lead Magnets: 2026 Guide": https://connectsafely.ai/articles/linkedin-lead-magnet-guide-2026
Best public prompt we found for this job
Quoted verbatim, its own punctuation included, from https://github.com/sergebulaev/linkedin-skills/blob/main/references/hook-formulas.md (section F6 "Comment-Gate Lead Magnet", line 178; repo 868 stars per api.github.com; 214 words).
## F6 — Comment-Gate Lead Magnet
**Reference:** 717-3,008 eng. Canonical draft: `06-comment-gate-leadmagnet.md`.
```
[Authority number: "We've helped creators publish 47,000+ posts in 14 months"]
[Pattern observation the authority earned]
So I turned that workflow into {N named items}. [Drop them into X, type one command, get the output.]
What's inside:
— [Item 1]
— [Item 2]
...
— [Item 12]
Free. No email wall. [Light scarcity — "48 hours only, I'll DM the link personally"]
Comment "{keyword}" below + connect with me and I'll send the bundle.
```
**Why:** Capped reach but huge DM conversion. Named bundle + real authority = 300-800 comments if the bundle is genuine.
**Warning:** This is engagement bait. Ship only when the weekly goal is list-building, not thought leadership. LinkedIn suppresses pure "comment X" posts.
**2026 status (gray zone, format-sensitive):** comment-gates still outperform link-in-body posts (they keep interaction on-platform), but keyword-only templated triggers are increasingly detected and suppressed. If shipped, the gate must sit inside a genuinely substantive post: real authority numbers, a named specific bundle, personalized framing, never a bare "comment X" line. For expert/technical audiences (developers, CTOs) the trust cost usually exceeds the reach gain; prefer F1-F5 there. Evidence: van der Blom Algorithm Insights 2026; LinkedIn's 2025 comment rate-limit policy; Social Media Today coverage of anti-pod enforcement.
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.
Foxera