LinkedIn Growth, Packaged as Markdown Prompts

What it does This repo is essentially a bundle of 11 SKILL.md prompt files—plus thin Python clients for optional data fetching and publishing—that plug into Claude Code, Codex, OpenClaw, and other agentic tools. It instructs the agent to draft LinkedIn posts, comments, replies, and content calendars using a library of 20 hook formulas and founder-specific angles, then strips out AI tells like em dashes and words such as “leverage” and “delve.” Every skill pauses for human approval before anything goes live.

The interesting bit The project treats prompt engineering as infrastructure. Most of the “code” is actually voice rules, hook templates, and audit checklists written in markdown, while the Python layer is just a minimal Apify reader and a Publora poster. The founder-specific layer is unusually concrete: it ships ten pre-built narrative angles (e.g., “audience of one,” “scarce-shots math”) mapped to engagement goals, which is more editorial strategy than software.

Key highlights

• 11 skills covering posts, comments, replies, audits, humanization, hook extraction, content planning, engagement monitoring, profile optimization, employee advocacy, and cross-platform repurposing
• “Humanizer” skill targets 2026 AI-detection patterns, including emoji-density scoring and multi-detector spread testing
• Optional Apify integration to read post bodies, comment threads, and engager lists without browser cookies
• Optional Publora integration for direct publishing to LinkedIn (and X, Threads, Instagram) from the terminal
• Founder mode with dedicated hook formulas and a content plan optimized for trust with investors and design partners, not raw impressions

Caveats

• The Python code is thin glue; the real value is in the markdown prompts, so don’t expect heavy automation logic
• Four skills require an Apify token to fetch LinkedIn data automatically; otherwise you are manually pasting URLs and text
• Publora’s free tier caps at 15 posts per month, and the README notes LinkedIn’s own API quirks (URL format mismatches, thread flattening bugs) that the publisher has to work around

Verdict Content creators, founders, and marketers who already live inside Claude Code or Codex will get the most mileage; if you are looking for a standalone SaaS dashboard or a fully autonomous bot, this is just a well-organized prompt pack with some Python duct tape.