phuaky XPoster
Autonomous agent for processing raw ideas into engagement-optimized X posts.
CRITICAL: NO THREADS. Only single medium-form posts (40-280 words).
Quick Start
- Clone this repo into
~/.claude/skills/XPoster/ - Copy
config.yaml.exampletoconfig.yamland edit - Run the Setup workflow:
"setup xposter" - Start drafting:
"/post"or"draft: your idea here"
Workflow Routing
| Workflow | Trigger | File |
|---|---|---|
| Setup | "setup xposter", "configure xposter" | Workflows/Setup.md |
| Draft | "/post", "/capture", "draft:", "tweet about", "post about", "turn this into a post" | Workflows/Draft.md |
| BatchProcess | "process input", "scan input", "batch process" | Workflows/BatchProcess.md |
| Validate | "validate", "score", "quality check" | Workflows/Validate.md |
| Organize | "organize", "move to ready/drafts" | Workflows/Organize.md |
| Schedule | "schedule", "queue for posting" | Workflows/Schedule.md |
| GenerateVisual | "generate image", "add visual" | Workflows/GenerateVisual.md |
| UpdateAlgorithm | "update algorithm", "fetch nikita" | Workflows/UpdateAlgorithm.md |
Tools
# Score a post inline
bun run Tools/ScoreTweet.ts "Your tweet"
# Score from a JSON file
bun run Tools/ScoreTweet.ts --file path/to/tweet.json
MCP Integration
Posting requires the X MCP server (@anthropic/mcp-x-server or equivalent):
mcp__x-server__create_tweet({ text: "content" })
mcp__x-server__reply_to_tweet({ tweet_id: "id", text: "reply" })
Reference Documentation
- Content & scoring rules:
Reference/ContentRules.md - X algorithm knowledge:
Reference/XAlgorithm.md - Post templates:
Reference/TweetFormats.md - Account setup:
Reference/AccountStructure.md
Examples
Example 1: Draft from conversation
User: "/post"
-> Invokes Draft workflow (conversation source)
-> Mines conversation for postable moments
-> Scores, iterates, returns ready draft
Example 2: Draft from idea
User: "draft: spent 3 hours debugging with Claude, switched to GPT and fixed it in 2 min"
-> Invokes Draft workflow (inline source)
-> Optimizes, scores, returns ready tweet
Example 3: Batch processing
User: "process input for myaccount"
-> Invokes BatchProcess workflow
-> Scans INPUT folder, analyzes each item
-> Scores all drafts, saves to OUTPUT/drafts/