NV
All skills
RedditLinkedInSocial sellingClaude CodeMIT · Open source

delta-engage

A Claude Code skill that surfaces high-fit posts to engage with — based on who actually needs what you do — and hands you a twice-weekly digest with a ready-to-edit comment for each. You always write the final reply and post it yourself; the skill never posts for you.

Get it on GitHub/delta-engage
Get started in 30 seconds

Install & run

1. Paste this into your terminal — it clones the skill into Claude Code and runs setup.

Install
git clone --depth 1 https://github.com/newan2001/delta-engage.git ~/.claude/skills/delta-engage && cd ~/.claude/skills/delta-engage && ./setup

2. Then run it inside Claude:

Run in Claude
/delta-engage

Requirements

  • Claude Code (with skills enabled)
  • Python 3 + requests (setup checks/installs this)
  • An Apify API token (BYOK) — free to get at apify.com
  • Optional: a free Reddit script app for richer engagement counts
What it does

Finds buyers, not competitors

Searches the pain your buyer voices — not your service category — then classifies every post as buyer, peer, kol, or noise, so a competitor's sales post never shows up as a lead.

Drafts the comment, in your voice

Each pick comes with a ready-to-edit reply, optionally voice-matched from your own public LinkedIn. Never fabricated experience, never auto-posted.

Reddit-safe by design

Bakes in current anti-shadowban practice: no link-spam, the 9:1 rule, per-subreddit norms, new-account ramps, and personalize-before-you-post.

Cookieless & zero account-risk

LinkedIn discovery uses only logged-out public actors — it never touches your session. You engage manually from your real account.

BYOK — bring your own key

Your own Apify token runs everything. No shared keys, no lock-in, and your config never lives in the repo.

How it works
1

Onboard

The first run reads your site and docs to draft your demand-focused ICP — you confirm it in a short walkthrough.

2

Fetch & dedup

Pulls fresh posts from Reddit and LinkedIn cookielessly, then normalizes and deduplicates them.

3

Classify & rank

Scores intent (buyer / peer / kol) against your goal, routing peers to a separate partnerships list instead of dropping them.

4

Draft & deliver

Writes a comment and angle for each pick and delivers a digest you can action in ~15 minutes — in-app, Slack, or Notion.

Free, MIT-licensed, yours to fork

Clone it, read the source, adapt it to your workflow. If it's useful, a star on GitHub helps others find it.