xoperator Customization Guide

English | 简体中文

Core Philosophy

xoperator is designed for high-quality replies, NOT suitable for:

Bulk marketing and spamming
Low-quality automated replies
Indiscriminate interactions

Suitable for you if you:

Have a clear persona and topic direction
Are willing to invest time in screening candidates and refining writing
Prioritize quality over quantity (better to skip than to spam)

Quick Customization Guide

1. Persona & Topic Positioning (Most Important, Required)

File Location: PROJECT.md §1 (Account Positioning & Quality Gates)

This is the foundation of your entire operations strategy, directly determining your content style and audience positioning.

Current Default Config

- Persona: First-person, young female tech practitioner; switches between English/Japanese following original tweet; natural, actionable tone.
- Content scope: Focus on hot topics that attract users and provide high-value insights; emphasize AI, tech, social phenomena, life philosophy. Avoid politics, economics, gossip, NSFW, unverifiable rumors, and avoid specialized technical theory outside AI (low traffic).

Customization Examples

Case 1: Web3 Founder Perspective

- Persona: First-person, Web3 founder; primarily English; pragmatic, critical thinking tone.
- Content scope: Focus on DeFi/DAO/NFT practical cases; crypto-economic mechanism design, community governance, tokenomics. Avoid pure speculation topics, airdrop hunting content, political stances.

Case 2: Japanese Native Creator

- Persona: First-person, Japanese tech observer; primarily Japanese, occasional English; gentle, nuanced tone.
- Content scope: Japanese tech industry, AI applications in Japan, East-West tech culture comparisons. Avoid politically sensitive topics, West-centric discussions.

Case 3: Product Manager Perspective

- Persona: First-person, B2B SaaS product manager; English; analytical, data-driven tone.
- Content scope: Product design methodology, user research, growth strategies, PLG models. Avoid pure technical implementation details, startup platitudes, unvalidated theories.

How to Modify

Open PROJECT.md
Search for "## 1. 账号定位与质量闸" or navigate to §1
Modify the Persona and Content scope items
Save file

Tip: These two lines will be used by PEERB to judge candidate suitability and as style guidance during writing.

2. Technical Threshold Adjustment (Optional)

File Location: CONFIG.json

These parameters control technical gates for candidate screening. Tools automatically read after modification.

Key Parameters

{
  "candidate_search": {
    "baseline": {
      "min_followers": 20000,       // Minimum author followers (hard floor)
      "max_age_hours": 12,          // Maximum tweet age (hours)
      "min_comments": 5,            // Minimum comments (avoid no interaction)
      "max_comments": 100,          // Maximum comments (avoid drowning)
      "min_views": 20000            // Minimum views (need exposure)
    },
    "sweep_strategy": {
      "follower_steps": [20000, 30000, 40000, 50000],  // Gradient sweep range
      "age_steps": [12, 10, 8, 6],                     // Time window tightening
      "views_steps": [20000, 50000, 80000]             // View count escalation
    }
  }
}

Customization Examples

Lower Thresholds (Get More Candidates):

{
  "baseline": {
    "min_followers": 10000,   // 20k → 10k
    "max_age_hours": 12,      // ⚠️ Fixed at 12h (hot tweet freshness is critical)
    "min_comments": 3,        // 5 → 3
    "max_comments": 100,      // Keep 100 (anti-drowning)
    "min_views": 10000        // 20k → 10k
  }
}

Raise Thresholds (Stricter Filtering):

{
  "baseline": {
    "min_followers": 50000,   // Focus on bigger accounts only
    "max_age_hours": 6,       // Only freshest tweets
    "min_comments": 10,       // Need some interaction
    "max_comments": 50,       // Stricter anti-drowning
    "min_views": 50000        // High exposure required
  }
}

⚠️ Note:

max_comments: 100 is critical for anti-drowning, recommend keeping
Gradient sweep tightens from baseline to sweep_strategy upper limits
Minimum candidate count floor_min_rows: 2, <2 triggers "sample too cold" warning

3. Quality Framework Adjustment (Advanced)

File Location: PROJECT.md §6.0 (Writing Principles section)

Default quality framework is relatively complex, including:

S4 Standards: Sharp / Surprising / Supported / Significant
Insight Wedges: Frame-Shift / Trade-off / Why-Now / Synthesis / Name-It / Exit
Hook / Proof / H2H: Three-question mandatory check

Simplified Version (For Beginners)

If too complex, can simplify to:

Open PROJECT.md
Find §6.0 "Writing Principles" section

Simplify focus to:

Only do three things:
1) Contextual response: Mention core content of original tweet
2) One real insight: Provide new angle or observation
3) Clean expression: ≤240 chars, no emoji, no CTA

Optional: Comment out §6.6 mandatory Hook/Proof/H2H checks

⚠️ Note: Quality framework is key to preventing mediocre output, simplify with caution.

4. Operations Rhythm Adjustment

File Location: PROJECT.md §3 (Heartbeat section) + CONFIG.json

Publication Frequency Limits

In PROJECT.md §0+ (One-Page Execution Card):

- microtopic 12h ≤2        # Same topic max 2 within 12h
- Same author 60min ≤1     # Same author max 1 reply per 60min
- Post interval ≥10min     # Minimum 10min between posts
- Rolling 60min ≤6 posts   # Max 6 per hour

Adjust these numbers as needed.

Heartbeat Rhythm

In PROJECT.md §3 (Heartbeat section):

Heartbeat = 30 minutes/beat, split into three 10-minute Time Boxes

Can adjust to faster or slower rhythm:

Fast: Heartbeat = 20 minutes (source 7min / consensus 6min / write 7min)
Slow: Heartbeat = 60 minutes (source 20min / consensus 20min / write 20min)

5. Content Type Ratio

File Location: PROJECT.md §4.2 (Content Inclination) and §10 (Default Parameters)

Current config:

Reply as main mode, Original requires explicit authorization

⚠️ Note: The system now focuses on Reply mode by default. Original posts require --allow-original flag.

6. Language Configuration

File Location: CONFIG.json

{
  "allow_lang": ["en", "ja"]  // Allowed languages
}

Can modify to:

["en"] - English only
["ja"] - Japanese only
["en", "ja", "zh"] - Add Chinese support (need sync changes in PROJECT.md language docs)

Advanced Customization

Influencer Pool Customization

File Location: INFLUENCER_POOL.json

Default includes 273 accounts (avg 400k followers, from influx project).

Custom Influencer Pool

Method 1: Derive from Existing Pool

# Edit INFLUENCER_POOL.jsonl (source data)
# Then regenerate
python3 tools/xop-influencer-derive.py

Method 2: Manual Edit

[
  {
    "handle": "@yourTargetAccount",
    "author_id": "123456789",
    "followers_count": 500000,
    "lang_primary": "en",
    "topic_tags": ["AI", "Tech"],
    "score": 90.0,
    "banned": false,
    "note": "Your note"
  }
]

Blacklist Management

Files: BANNED_HANDLES.txt and BANNED_PHRASES.txt

BANNED_HANDLES.txt: One @handle per line, these accounts will be filtered
BANNED_PHRASES.txt: One phrase per line (case-insensitive), tweets containing these will be filtered

Quick add tool:

python3 tools/xop-ban.py --handle @spammer_account

Writing Style Customization

File Location: PROJECT.md §6.3 (Writing Examples section)

Provides English/Japanese writing examples reflecting specific style features:

Concise and sharp
Technically oriented
Observational rather than preachy

You can:

Replace these examples with your own style
Add more examples for reference
Use these examples as standards during PEERB review

Tip: These examples are mainly for human/Aux reference, not hard templates.

Parts NOT Recommended for Customization

Following are core quality assurance mechanisms, not recommended for modification:

🚫 Don't Modify

§0 RED FLAGS (Real-Ops Pledge section)
- Comments ≥100 must stop - prevent drowning
- Prohibit reply-of-reply - maintain content quality
- Publication time >12h baseline - ensure timeliness
Provenance Validation Logic
- All candidates must come from real fetches
- Prohibit manual table filling - prevent data fabrication
- xop-lint.py validation rules
Data Integrity Checks
- 10 slice file completeness requirement
- DQF (Data Quality Firewall)
- Sample statistics audit

These are project's core design principles. Modifications may lead to quality degradation or system failure.

Customization Checklist

After customization, run these checks to ensure config correctness:

# 1. Syntax check
make test

# 2. Config diagnostics
make diagnose

# 3. Complete system diagnostics
python3 tools/xop-preflight.py --diagnose

# 4. Lint check (if candidate lists exist)
python3 tools/xop-lint.py

Real Case: Complete Customization Flow

Case: From Default to "Product Manager + SaaS Domain"

Step 1: Modify Persona (PROJECT.md §1)

- Persona: First-person, B2B SaaS product manager; English; analytical, data-driven tone.
- Content scope: Product design, user research, PLG growth, SaaS metrics analysis. Avoid pure technical implementation, unvalidated theories, startup platitudes.

Step 2: Adjust Influencer Pool (Focus on SaaS Domain)

Edit INFLUENCER_POOL.json, keep/add SaaS-related accounts:

@lennysan (Lenny's Newsletter)
@joulee (Julie Zhuo)
@shreyas (Shreyas Doshi)
@gokulrajaram (Gokul Rajaram)
etc.

Step 3: Adjust Thresholds (Optional, keep defaults OK)

{
  "baseline": {
    "min_followers": 30000,  // SaaS circle relatively niche, can raise slightly
    "max_age_hours": 12,
    "min_comments": 5,
    "max_comments": 100,
    "min_views": 20000
  }
}

Step 4: Update Writing Examples (PROJECT.md §6.3)

Add product manager style examples:

- English Example (Product):
  "Retention isn't a feature problem—it's an onboarding promise problem. If your D7 falls off a cliff, audit what you showed vs. what you delivered in Week 1."

Step 5: Verify

make diagnose
python3 tools/xop-setup.py --check

Common Customization Q&A

Q: Want Chinese operations, how to modify?

A: Need multiple modifications:

CONFIG.json: "allow_lang": ["zh"]
PROJECT.md §1: Modify language docs in persona
PROJECT.md §6.1: Add Chinese lint rules
Prepare Chinese Influencer pool

Q: How to completely disable quality framework, just do simple replies?

A: Not recommended, but if necessary:

Comment out PROJECT.md §6.6 check rules
Simplify §6.0 to basic "contextual+insight+clean"
Risk: Easy to produce low-quality content

Q: Can adjust to 10 posts per hour?

A: Yes, but not recommended:

Modify PROJECT.md §0+ rate limit rules
Risk: High-frequency posting easily perceived as bot, affecting account safety

Q: Config file format wrong, what to do?

A: Run diagnostic tool:

make diagnose  # Will check JSON format and key fields

Getting Help

Documentation: See PROJECT.md (Operations Constitution) and README.md (Quick Start)
Issues: Submit GitHub Issue
Community: Currently in early stage, no dedicated community yet

Final Reminder:

xoperator's core value is "better to skip than spam" quality control. When customizing, recommend maintaining this principle:

✅ Adjust topic direction based on your domain and persona
✅ Adjust thresholds based on target audience
⚠️ Carefully modify quality framework
❌ Don't lower standards for "posting more"

High-quality few posts >> Low-quality many posts

FilesExpand file tree

CUSTOMIZATION.md

Latest commit

History

CUSTOMIZATION.md

File metadata and controls

xoperator Customization Guide

Core Philosophy

Quick Customization Guide

1. Persona & Topic Positioning (Most Important, Required)

Current Default Config

Customization Examples

How to Modify

2. Technical Threshold Adjustment (Optional)

Key Parameters

Customization Examples

3. Quality Framework Adjustment (Advanced)

Simplified Version (For Beginners)

4. Operations Rhythm Adjustment

Publication Frequency Limits

Heartbeat Rhythm

5. Content Type Ratio

6. Language Configuration

Advanced Customization

Influencer Pool Customization

Custom Influencer Pool

Blacklist Management

Writing Style Customization

Parts NOT Recommended for Customization

🚫 Don't Modify

Customization Checklist

Real Case: Complete Customization Flow

Case: From Default to "Product Manager + SaaS Domain"

Step 1: Modify Persona (PROJECT.md §1)

Step 2: Adjust Influencer Pool (Focus on SaaS Domain)

Step 3: Adjust Thresholds (Optional, keep defaults OK)

Step 4: Update Writing Examples (PROJECT.md §6.3)

Step 5: Verify

Common Customization Q&A

Q: Want Chinese operations, how to modify?

Q: How to completely disable quality framework, just do simple replies?

Q: Can adjust to 10 posts per hour?

Q: Config file format wrong, what to do?

Getting Help