LinkedIn Copywriter

The linkedin-copywriter skill is part of the gtm-skills bundle — a Claude Code skill catalog for GTM workflows. It’s a focused instruction set that loads on demand when an agent is asked to draft LinkedIn content. It uses the gtm-tools LinkedIn tools for voice ingestion + context fetching, then drafts and scrubs against an anti-AI pattern list built from the top LinkedIn-ghostwriter skills on the market.

Installation

Install just the LinkedIn Copywriter skill:

$
npx skills add arnaudjnn/gtm-skills --skill linkedin-copywriter

Or install the whole gtm-skills catalog (outbound, signals, linkedin-copywriter, reddit-community-manager, setup):

$
npx skills add arnaudjnn/gtm-skills

Both work because arnaudjnn/gtm-skills follows the anthropics/skills layout — each skill lives at skills/<name>/SKILL.md and is addressable by name.

You can also install via Claude Code’s plugin marketplace (uses .claude-plugin/marketplace.json):

/plugin marketplace add arnaudjnn/gtm-skills
/plugin install gtm-skills@gtm-skills

Set the API key once before invoking any tool:

$
export GTM_TOOLS_API_KEY="<your-api-key>"
$
# Get one via `get_api_key` at https://gtm-tools.sh,
$
# or `gtm-tools admin login` from the CLI.

Summary

When the linkedin-copywriter sub-skill loads, it teaches the agent:

• Four modes — post / comment / invitation / dm — each with its own char budget, structure, and tool fan-out. The skill picks the mode from the user’s request and refuses to draft if ambiguous.
• Voice ingestion — pulls the user’s last 15–30 LinkedIn posts via list_user_posts to learn their register before drafting. No copying phrases; capturing rhythm, opener patterns, content pillars, formatting habits.
• Reaction-Mode Gate (for comments) — extracts the original poster’s claims verbatim and asks the user to confirm which one(s) to react to before drafting. Catches misreadings before they become public, downvotable comments.
• Anti-AI scrub — mandatory pass against ~55 banned corporate-tells (leverage, utilize, robust, paradigm, synergy, streamline, comprehensive, pivotal, transformative, cutting-edge, ecosystem, holistic + intensifiers like deeply, truly, fundamentally) plus em-dash ban, rule-of-three ban, sycophancy ban, hedge-stacking ban. Each failure: rewrite.
• Char-limit guards — 3,000 hard / 1,800–2,800 sweet for posts; 1,250 hard / 200–350 sweet for comments; 300 hard for invitations; ~800 target for DMs.
• Draft-then-paste by default — posts and comments output as copy-paste-ready text; invitations and DMs auto-send via send_linkedin_invitation / send_linkedin_message only after explicit user approval (“ship it” / “send it”, not “looks good”).

The skill is single-purpose by design: it doesn’t try to teach the whole gtm-tools surface, just LinkedIn copywriting done correctly. For prospecting / search workflows, your agent uses the LinkedIn Tools reference directly.

Skill

The skill source lives in the arnaudjnn/gtm-skills public repo and is listed on the open-skills directory: