This project provides an AI-powered Claude Code skill designed to automate social media content creation and posting across Facebook, Instagram, Threads, and X. It addresses the challenge of maintaining a consistent and engaging online presence by learning a user's unique voice, generating a 14-day content calendar using proven viral formulas, and automating the posting process. The target audience includes content creators, small businesses, and individuals seeking to scale their social media reach efficiently, offering the benefit of significant time savings and data-driven content strategy.

How It Works

The skill leverages Claude Code to analyze a user's past Facebook posts (minimum 20) to extract linguistic patterns, vocabulary, and emoji usage, creating a style_profile.md. It then employs seven distinct "viral formulas," reverse-engineered from successful content by creators with under 5,000 followers, to populate a 14-day content_plan.md based on user-defined goals. Posting is automated via Chrome MCP, simulating user actions in a browser, with a mandatory confirmation step before content is published to the selected platforms. Performance metrics are tracked, enabling iterative optimization of content strategies.

Quick Start & Requirements

• Primary install / run command:
  1. Clone the repository: git clone https://github.com/Hao0321/claude-skill-social-post.git
  2. Copy the social-post directory to your Claude skills path (e.g., ~/.claude/skills/social-post on macOS/Linux).
  3. Rename style_profile.example.md to style_profile.md and content_plan.example.md to content_plan.md within the skill's directory.
• Non-default prerequisites and dependencies: Claude Code (macOS, Windows, Linux), Chrome browser, a Facebook personal account with at least 20 public posts for style learning, and Chrome MCP installed.
• Usage: Interact with Claude Code using conversational Chinese commands (e.g., "幫我學 FB 風格" to learn style, "幫我排社群內容日曆" to plan content, "今天發一篇" to post).
• Links: GitHub repository (implied by clone URL), Claude Code Taiwan discussion group (Line).

Highlighted Details

• "Mega-viral validated" with metrics like 72K reach, 358 likes, 443 comments, and significant community growth (+700 members) on a single post.
• Includes detailed post-mortem analyses of less successful posts ("flop postmortem") for learning.
• Employs 7 specific viral content formulas (F1-F7) tailored for smaller creator accounts, distinct from generic advice.
• Version 1.0.1 (2026-06-10) introduced "real voice + keyword CTA" features, demonstrating up to 10x audience increase and record comment counts.
• The project has undergone rapid iteration (v0.6 to v1.0.1) with frequent updates incorporating real-world data, algorithmic insights, and new content archetypes.

Maintenance & Community

The project is actively maintained by the original author, Hao (駱君昊), with a clear history of iterative development and bug fixing, including the use of a sister skill (claude-skill-code-cleanup) for repository audits. Community engagement is fostered through a Line discussion group and GitHub issues/pull requests.

Licensing & Compatibility

• License type: MIT License.
• Compatibility notes: Permissive for modification, usage, and commercial application, provided the LICENSE file is retained and original authorship is credited.

Limitations & Caveats

Facebook's UI (DOM) is subject to frequent changes, which may occasionally break the automation, although fallback mechanisms are in place. Instagram posting requires an image, and text-only posts are automatically adapted. Threads desktop lacks account-switching functionality, necessitating mobile use for multi-account management. Public posts are irreversible, and the skill enforces a user confirmation step before publishing. Effective style learning requires a minimum of 20 past public Facebook posts.