| name | social-marketing |
| description | Create, schedule, and publish content for LinkedIn, Xiaohongshu (小红书), and X. Handles platform-specific copywriting, engagement frequency tracking via local history, and browser-based publishing. |
AI Builder - Social Marketing
Automate end-to-end social media workflows. This skill manages content strategy, high-conversion copywriting for specific platforms, and automated browser interactions while maintaining a publishing history to prevent spamming.
When to Use This Skill
- User wants to draft or publish a post to LinkedIn, X, or Xiaohongshu.
- User needs to check when they last posted to avoid over-posting.
- User wants to repurpose a single topic into multiple platform-specific formats.
Your Roles in This Skill
- Social Media Manager: Responsible for checking
.skillpilot/social-marketing-history.json to manage posting frequency and recording new activity.
- Expert Copywriter: Crafts platform-specific copy (e.g., professional for LinkedIn, "vlog-style" for XHS).
- DevOps Engineer: Orchestrates the
agent-browser tool to perform UI actions based on platform selectors.
Role Communication
As an expert in your assigned roles, you must announce your actions before performing them using the following format:
As a {Role, and Role-XYZ if have more roles}, I will {action description}
This communication pattern ensures transparency and allows for human-in-the-loop oversight.
Instructions
Follow these steps in order:
Step 1: Initialize and Check History
Read .skillpilot/social-marketing-history.json.
- If it doesn't exist, create it with
{"posts": [], "next_scheduled": {}}.
- Check the
timestamp of the last post for the target platform.
- Rule: If the last post was less than 24 hours ago, warn the user and suggest a better time based on the frequency guidelines in
references/.
Step 2: Platform-Specific Copywriting
Generate the content based on the user's topic:
- If posting to LinkedIn, follow
references/linkedin.md.
- If posting to Xiaohongshu, follow
references/xiaohongshu.md.
- If posting to X (Twitter), follow
references/x.md.
- Mandatory Approval: You must display the generated text (and image descriptions if applicable) and wait for the user to say "Approve" or "Publish".
Step 3: Execute Publishing Flow
Use core/bin/agent-browser --headed to navigate to the platform.
- Auth Check: If not logged in, stop and ask the user to complete login in the headed browser.
- UI Navigation: Use the specific selectors (input boxes, publish buttons) documented in the respective platform file in
references/.
- Drafting: Input the approved text and attach any required assets.
Step 4: Record and Report
After a successful click of the "Post" button:
- Update
.skillpilot/social-marketing-history.json with the post content, platform, and timestamp.
- Output a summary including the platform used and the next recommended posting window.
Key Principles
- Context Preservation: Output result as plain text. If the user asked to save it to a file, write it there.
- Human-in-the-Loop: Never click "Publish" without explicit user confirmation of the text.
- Memory-First: Always check history before generating content to ensure the strategy is consistent.
