Autonomous Hands
LibreFang's core innovation — 7 pre-built autonomous capability packages that run independently and execute on schedule without human intervention.
Tip: Hands can be configured with cron expressions for scheduled execution, and also support event-triggered mode.
Overview
Hands are LibreFang's unique feature — they are autonomous agents that can:
- Execute tasks automatically on schedule
- Operate across multiple platforms
- Handle complex multi-step workflows
- Make decisions without human intervention
7 Built-in Hands
| Hand | Function | Autonomy Level |
|---|---|---|
| Clip | YouTube video download, clipping, subtitle generation, vertical short video publishing | High |
| Lead | Daily lead discovery, enrichment, 0-100 scoring, deduplication | High |
| Collector | OSINT intelligence collection, change detection, sentiment tracking, knowledge graph | High |
| Predictor | Superforecasting engine, confidence intervals, contrarian patterns | High |
| Researcher | Deep autonomous research, CRAAP credibility assessment, APA citation | High |
| Autonomous X/Twitter account management, 7 content formats, approval queue | High | |
| Browser | Web automation, Playwright, mandatory purchase approval gate | Medium |
Clip Hand
YouTube video automation Hand.
Features
- Video download (multiple formats and quality options)
- Automatic clipping and trimming
- Subtitle generation and translation
- Vertical short video adaptation (9:16)
- Auto-publish to social platforms
Configuration
[clip]
youtube_quality = "1080p"
auto_subtitle = true
vertical_format = true
publish_targets = ["tiktok", "youtube_shorts"]
CLI Commands
# Activate Clip Hand
librefang hand activate clip
# Download video
librefang hand clip download "https://youtube.com/watch?v=..."
# Clip video
librefang hand clip clip --start 60 --end 120 --input video.mp4
# Generate subtitles
librefang hand clip subtitle --lang en video.mp4
Lead Hand
Daily lead discovery and scoring Hand.
Features
- Automatic lead discovery
- Company information enrichment (size, industry, contact info)
- 0-100 quality scoring
- Deduplication and merging
- CRM integration
Configuration
[lead]
icp = "companies in AI space, >100 employees, Series A+"
search_sources = ["linkedin", "crunchbase", "hunter"]
score_criteria = ["funding", "tech_stack", "team_size"]
deduplicate = true
crm_integration = "salesforce"
CLI Commands
# Activate Lead Hand
librefang hand activate lead
# Manual lead discovery
librefang hand lead discover --icp "B2B SaaS, >50 employees"
# View leads list
librefang hand lead list
# Export to CSV
librefang hand lead export --format csv
Collector Hand
OSINT intelligence collection Hand.
Features
- Continuous target monitoring
- Change detection and notifications
- Sentiment analysis
- Knowledge graph construction
- Multi-source data aggregation
Configuration
[collector]
targets = ["competitor.com", "industry-news.com"]
monitor_interval = "1h"
sentiment_analysis = true
knowledge_graph = true
alert_channels = ["telegram", "email"]
CLI Commands
# Activate Collector Hand
librefang hand activate collector
# Add monitoring target
librefang hand collector add-target "https://example.com"
# View collected data
librefang hand collector list
# Export knowledge graph
librefang hand collector export-graph
Predictor Hand
Superforecasting engine Hand.
Features
- Data-based predictions
- Confidence interval calculation
- Contrarian pattern recognition
- Prediction verification and calibration
- Trend analysis
Configuration
[predictor]
prediction_horizon = "30d"
confidence_level = 0.95
contrarian_mode = true
calibration_history = true
CLI Commands
# Activate Predictor Hand
librefang hand activate predictor
# Create prediction
librefang hand predictor predict --question "Will Q4 revenue exceed $1M?"
# View prediction history
librefang hand predictor history
# Verify prediction
librefang hand predictor verify --prediction-id <id> --actual <value>
Researcher Hand
Deep autonomous research Hand.
Features
- Automatic topic research
- Multi-source information collection
- CRAAP credibility assessment
- APA format citations
- Structured report generation
Configuration
[scientist]
sources = ["academic", "news", "reports"]
citation_format = "apa"
max_sources = 20
credibility_threshold = 0.7
report_format = "markdown"
CLI Commands
# Activate Researcher Hand
librefang hand activate researcher
# Execute research
librefang hand researcher search "impact of AI on software development"
# View research history
librefang hand researcher history
# Generate report
librefang hand researcher report --id <research-id>
Twitter Hand
Autonomous social media management Hand.
Features
- Automatic content creation
- 7 content formats
- Approval queue
- Auto-publishing
- Interaction replies
Content Formats
- Thread - Long tweet threads
- Hot Take - Hot takes
- Question - Community questions
- Fact - Fact sharing
- Quote - Quote tweets
- Announcement - Product announcements
- Engagement - Engagement tweets
Configuration
[twitter]
content_formats = ["thread", "hot_take", "question"]
approval_queue = true
auto_schedule = true
posting_times = ["9:00", "12:00", "18:00"]
CLI Commands
# Activate Twitter Hand
librefang hand activate twitter
# Create content
librefang hand twitter create --format thread --topic "AI trends"
# View approval queue
librefang hand twitter queue
# Approve tweet
librefang hand twitter approve <tweet-id>
Browser Hand
Web automation Hand.
Features
- Playwright-driven
- Form filling
- Click interactions
- Multi-step workflows
- Purchase approval gate
Configuration
[browser]
headless = false
purchase_approval = true
screenshot_on_error = true
timeout = 30000
CLI Commands
# Activate Browser Hand
librefang hand activate browser
# Execute automation task
librefang hand browser run --workflow "checkout-flow"
# Record new workflow
librefang hand browser record --name "my-workflow"
# List workflows
librefang hand browser list
Hand Management
CLI Commands
# List all Hands
librefang hand list
# Activate Hand
librefang hand activate <hand-name>
# Pause Hand
librefang hand pause <hand-name>
# View status
librefang hand status <hand-name>
# View logs
librefang hand logs <hand-name>
# Configure Hand
librefang hand config <hand-name> --key <value>
Hand Status
| Status | Description |
|---|---|
| Active | Running, executing on schedule |
| Paused | Paused, can be resumed |
| Error | Error occurred, intervention needed |
| Pending | Waiting for approval |
Creating Custom Hands
Hand.toml Structure
name = "my-hand"
version = "1.0.0"
description = "Custom autonomous hand"
[schedule]
cron = "0 9 * * *"
timezone = "UTC"
[capabilities]
tools = ["web_fetch", "file_write"]
network = ["*"]
[config]
max_results = 100
timeout = 3600
[approval]
required = false
channels = ["telegram"]
Implementation Code
# hands/my-hand/main.py
async def execute(context: HandContext) -> HandResult:
"""Execute the hand's main logic"""
# Fetch data
data = await context.tools.web_fetch("https://api.example.com/data")
# Process data
results = process(data)
# Store results
await context.memory.store("hand_results", results)
return HandResult(
success=True,
output=results,
metadata={"count": len(results)}
)
Schedule Configuration
Cron Expressions
[schedule]
# Every day at 9 AM
cron = "0 9 * * *"
# Weekdays at 9 AM
cron = "0 9 * * 1-5"
# Every hour
cron = "0 * * * *"
# Every Monday
cron = "0 9 * * 1"
Timezone
[schedule]
timezone = "Asia/Shanghai"
Best Practices
- Start Simple - Use a single Hand first
- Set Up Monitoring - Regularly check Hand status
- Configure Alerts - Get timely error notifications
- Approval Workflow - Sensitive operations need approval
- Log Review - Regularly review execution logs
Troubleshooting
Hand Not Executing
# Check Hand status
librefang hand status <hand-name>
# View detailed logs
librefang hand logs <hand-name> --level debug
# Manually trigger execution
librefang hand run <hand-name>
Error Handling
# View error details
librefang hand error <hand-name> --last
# Reset Hand
librefang hand reset <hand-name>