skills/nirholas/xactions/algorithm-cultivation

algorithm-cultivation

SKILL.md

Algorithm Cultivation & Thought Leadership

Train your X/Twitter algorithm for a specific niche. Three approaches:

  1. Browser script -- paste into DevTools console for manual sessions
  2. CLI + Persona Engine -- create personas and run from the command line
  3. 24/7 Algorithm Builder -- headless Puppeteer + LLM running continuously

Quick Reference

Goal Solution
Create a persona (CLI) xactions persona create
Run 24/7 with LLM (CLI) xactions persona run <id>
Check persona status xactions persona status <id>
Browser console (with core.js) src/automation/algorithmBuilder.js
Browser console (standalone) scripts/thoughtLeaderCultivator.js
Browser console (algorithm trainer) src/automation/algorithmTrainer.js
Persona Engine (Node.js module) src/personaEngine.js
Algorithm Builder (Node.js module) src/algorithmBuilder.js

Core Concepts

  • Persona -- identity config: niche, activity pattern, engagement strategy, topics
  • Session -- one period of activity (search, browse, engage, post)
  • Strategy -- engagement limits (aggressive/moderate/conservative/thoughtleader)
  • Activity pattern -- human-like schedule (night-owl/early-bird/nine-to-five/always-on/weekend-warrior)

Algorithm Builder -- src/algorithmBuilder.js

24/7 headless automation: Puppeteer + stealth + OpenRouter LLM.

import { startAlgorithmBuilder } from './algorithmBuilder.js';

await startAlgorithmBuilder({
  personaId: 'persona_1234',
  authToken: 'your_auth_token',
  headless: true,
  dryRun: false,
  maxSessions: 0, // 0 = infinite
});

Requires OPENROUTER_API_KEY env var for LLM-generated comments and posts.

Algorithm Trainer -- src/automation/algorithmTrainer.js

Browser console script for manual training sessions. Requires src/automation/core.js pasted first.

Training Phases (cycles through all 8)

  1. Search top tweets for niche keywords
  2. Search latest tweets for niche keywords
  3. Follow people from search results
  4. Engage with home feed (like/reply)
  5. Visit influencer profiles
  6. Browse random profiles in niche
  7. Explore page browsing
  8. Idle dwell time (human-like pauses)

Controls

  • stopTrainer() -- Stop training
  • trainerStatus() -- Current phase, actions taken, rate limits
  • trainerReset() -- Reset counters

Intensity Presets

Preset Actions/hour Daily cap
chill 10-15 100
normal 20-30 300
active 40-60 500

Strategy Guide

Fresh account (week 1-2)

  1. Create a persona with xactions persona create or configure algorithmTrainer manually
  2. Use conservative/chill intensity -- X flags aggressive new accounts
  3. Focus on phases 1-2 (search) and 7 (explore) to signal interests
  4. Follow 5-10 niche accounts per day manually
  5. Like 20-30 niche tweets per day

Established account pivoting niches

  1. Aggressively engage with new niche content for 1-2 weeks
  2. Use src/automation/algorithmTrainer.js on active intensity
  3. Unfollow accounts from the old niche gradually with src/automation/smartUnfollow.js
  4. The algorithm typically adjusts within 3-7 days of consistent signals

Running 24/7 with LLM

  1. Set OPENROUTER_API_KEY for AI-generated replies
  2. xactions persona create -- configure niche, strategy, schedule
  3. xactions persona run <id> -- starts headless Puppeteer session
  4. Monitor: xactions persona status <id>
  5. Cost estimate: ~$0.50-2.00/day depending on model and activity level

Environment Variables

Variable Purpose
OPENROUTER_API_KEY Required for LLM-generated comments and posts
XACTIONS_SESSION_COOKIE X auth token (alternative to --token flag)

Detailed References

Load these on demand for deeper context:

  • Persona Engine details -- skills/algorithm-cultivation/references/persona-engine.md
  • Browser scripts -- skills/algorithm-cultivation/references/browser-scripts.md
  • Algorithm internals -- skills/algorithm-cultivation/references/algorithm-internals.md
  • Research: algorithm internals -- docs/research/algorithm-cultivation.md
  • Research: LLM architecture -- docs/research/llm-powered-thought-leader.md

Notes

  • Browser scripts require navigating to x.com first
  • algorithmBuilder.js and algorithmTrainer.js require pasting src/automation/core.js first
  • thoughtLeaderCultivator.js is standalone (no dependencies)
  • Default engagement strategy: 1-3s delays between actions
  • Fresh accounts should start with conservative strategy for 1-2 weeks
Weekly Installs
4
GitHub Stars
108
First Seen
Feb 28, 2026
Installed on
openclaw4
gemini-cli4
github-copilot4
codex4
kimi-cli4
cursor4