LinkedIn Skills for Claude Code and Codex
11 Claude Code and Codex skills that write human-sounding LinkedIn posts, comments, and replies — all from your terminal.
Evidence shows: skills default to drafting only, with user approval required before publishing (README emphasizes 'waits for your approval'); API tokens are read from environment variables or .env, which is gitignored; SECURITY.md explicitly states no hardcoded credentials and discloses an optional DIY backend (LINKEDIN_SKILLS_CUSTOM_POSTER) that can lead to code execution but is unset by default. Data flow transparency: README explains Apify read, Publora publish, and Pixfaro image generation flows. External effects: publishing requires explicit user approval, and Publora free tier limits. Rollback: README mentions canceling scheduled posts in Publora dashboard, but no systematic rollback mechanism. Source attribution: README clearly attributes to Creative Content Crafts and author. Deductions: dependency security only claimed via GitHub Action scanner without specific results; rollback limited; publisher unverified.
Evidence shows: README is internally consistent, skill list matches descriptions; dependencies are only requests and python-dotenv, both common; failure messages provided in Troubleshooting table (e.g., 401, 400 errors). Deductions: no actual tests or run evidence, static review cannot verify self-consistency.
Evidence shows: README clearly defines target audience (founders, B2B SaaS, etc.) and scenarios (posting, commenting, auditing, etc.); capability boundaries via skill descriptions and 'Not for X' disambiguation; trigger precision via automatic skill activation; environment fit covers multiple runtimes (Claude Code, Codex, OpenClaw, etc.). Deductions: no actual trigger test evidence.
Evidence shows: clear information architecture (skills/, lib/, references/, etc.); detailed installation instructions for multiple runtimes; stable naming (skill names fixed); examples and FAQ in README; known limitations mentioned in SECURITY.md and README (e.g., DIY backend risk); MIT license explicit; maintenance responsibility via SECURITY.md and contribution guidelines. Deductions: versioning/changelog not explicitly provided (only release badge); maintenance responsibility does not specify update frequency.
Evidence shows: high output usability (drafts can be copy-pasted or published via Publora); clear marginal value (saves time, improves content quality); reasonable cost-benefit (free tiers and low-cost APIs). Deductions: no actual usage data or user feedback.
Evidence shows: claims in README (e.g., '20 proven hook formulas') lack specific sources or verification; cross-source corroboration limited (only mentions external tools like AuthoredUp); facts and inferences not clearly separated. Deductions: lack of traceable evidence chain.
- Publisher is unverified and identity unknown; exercise caution.
- Optional DIY backend (LINKEDIN_SKILLS_CUSTOM_POSTER) if set could lead to code execution; protect that variable strictly.
- Dependency security only claimed via scanner without specific results; recommend independent dependency review.
- Marketing claims in README (e.g., '20 proven hook formulas') lack verifiable sources; verify independently.
What does this agent do, and when should you use it?
LinkedIn Skills for Claude Code and Codex is an open-source (MIT) bundle of 11 skills that help marketers create, optimize, and plan LinkedIn content with a voice that feels human. Built around an approve-before-publish flow, the skills draft content and wait for your go-ahead, so they're safe to use with Claude Code, Codex CLI, claude.ai, OpenClaw, and Hermes Agent (via SKILL.md). The bundle covers post writing (using 2026 hook formulas), comment drafting, reply handling, post auditing, AI-tell removal (humanizer), hook extraction, content planning, engagement monitoring, profile optimization, employee advocacy, and repurposing. It also includes a founder-focused angle library and cross-referencing files. Optional integrations with Apify (read LinkedIn posts, comments, engagers), Publora (auto-publish to LinkedIn), and Pixfaro (illustration generation) extend its capabilities.
Provides 11 skills: Post Writer (16 hook formulas + founders-edition angles), Comment Drafter (drafts comments from URL), Reply Handler (handles LinkedIn's 2-level thread flattening), Post Audit (checks drafts against 2026 algorithm rules and AI-detection patterns), Humanizer (removes AI tells, caps em dashes), Hook Extractor (reverse-engineers viral hooks), Content Planner (creates 7-day plans), Engagement Monitor (tracks replies, likers/commenters), Profile Optimizer (rewrites headline, About, Featured, Experience), Employee Advocacy (plans team LinkedIn program), Repurposer (turns content from other platforms into native LinkedIn posts). Each skill shows a draft first and requires approval before posting. Reads/writes via optional integrations: Apify (reads posts/comments/engagers via APIFY_TOKEN), Publora (publishes via PUBLORA_API_KEY + LINKEDIN_PLATFORM_ID), Pixfaro (generates images via PIXFARO_TOKEN).
- A B2B SaaS founder wants to post consistently on LinkedIn without sounding robotic, using the Post Writer to draft viral-ready posts and approve before publishing.
- A marketing manager needs to engage daily with prospects, using Comment Drafter to write thoughtful comments on relevant posts.
- A content creator wants to audit drafts for AI-tells to maintain authenticity, using the Post Audit and Humanizer skills before hitting publish.
- A personal brand builder wants to optimize their LinkedIn profile to attract inbound leads, using Profile Optimizer to rewrite headline, About, and Experience.
- A team leader wants to implement an employee advocacy program, using Employee Advocacy to plan a 14-day launch and governance.
- A repurposing workflow for a podcaster who wants to turn a YouTube video into a LinkedIn post, using Repurposer to adapt the content to LinkedIn's style and length.
What are this agent's strengths and limitations?
- Comprehensive skill set covering the entire LinkedIn workflow: create, audit, engage, and plan.
- Strong emphasis on maintaining a human voice, with a dedicated humanizer and 2026 hook formulas.
- Approve-before-publish flow gives users full control, reducing risk of accidental posts.
- Founder-specific angles and content plans tailored for high-stakes audiences like investors and hires.
- Optional integrations (Apify, Publora, Pixfaro) extend capabilities without blocking core functions.
- Open source and MIT licensed, with a community for additional skills.
- Best experience is on Claude Code and Codex; other platforms (Claude Desktop, OpenClaw) require manual setup or feature tradeoffs.
- Requires external services for full functionality: Apify token and Publora API key, both with free tiers but limited quotas.
- No guarantees against AI detection; humanizer improves style but not proactive detection.
- Network and API dependencies may increase operational complexity.
- Skills are designed for LinkedIn only, lacking support for other networks beyond repurposing.
- Some advanced features (e.g., employee advocacy) may be overkill for individual creators.
How do you install or deploy this agent?
For Codex CLI: run codex plugin marketplace add sergebulaev/linkedin-skills and codex plugin add linkedin-skills@linkedin-skills. For claude.ai (web): open https://claude.ai/code, go to Skills, click Add from GitHub, paste sergebulaev/linkedin-skills. For Claude Desktop: open Customize, click + next to Personal plugins, Create plugin, Add marketplace, choose Add from a repository, paste the repo, install. For Claude Code (CLI/IDE): in conversation run /plugin marketplace add sergebulaev/linkedin-skills then /plugin install linkedin-skills@linkedin-skills. For OpenClaw: clone the repo into working directory and add a system prompt pointing to skills/*/SKILL.md and lib/ modules. For Hermes Agent: clone the repo into ~/.hermes/skills/linkedin-skills.
How do you use this agent?
After installation, simply ask Claude Code or Codex for help with LinkedIn, e.g., 'Write me a LinkedIn post about AI replacing traditional agencies.' The appropriate skill activates, drafts content, and waits for your approval. For enhanced features, create a .env file in the linkedin-skills folder with APIFY_TOKEN, PUBLORA_API_KEY, and LINKEDIN_PLATFORM_ID as needed. To auto-post, configure Publora by signing up at app.publora.com, connecting your LinkedIn, and copying the platform ID and API key into .env. Test with a scheduled post. The skills handle LinkedIn's URL formats and thread flattening automatically.
How does this agent compare with similar options?
Part of a family of social-media skill bundles for Claude Code and Codex, with parallel repositories for X (Twitter), Instagram, YouTube, TikTok, Threads, and Facebook, all sharing the same voice engine and flow.
FAQ
Do I need to know how to code to use these skills?
What does the free tier cover?
How do I integrate with Publora to auto-post?
.env file in the skills folder.What if the skills don't activate?
/plugin install linkedin-skills@linkedin-skills.