Skip to content

Creating Custom Skills

Custom skills extend claudefi's capabilities with reusable prompts and workflows.

Skill Structure

A skill is a markdown file in .claude/skills/user/:

markdown
# Skill Name

Brief description of what this skill does.

## Trigger

/command-name

## Arguments

- `arg1` - Description (required)
- `arg2` - Description (optional, default: value)

## Instructions

Detailed instructions for Claude when this skill is invoked.

## Examples

Example usage and expected behavior.

Creating Your First Skill

1. Create the File

bash
mkdir -p .claude/skills/user
touch .claude/skills/user/analyze-pool.md

2. Write the Skill

markdown
# Analyze Pool

Deep analysis of a Meteora DLMM pool for potential LP entry.

## Trigger

/analyze-pool

## Arguments

- `address` - Pool address to analyze (required)

## Instructions

When the user invokes /analyze-pool:

1. Fetch pool data using the Meteora API
2. Calculate key metrics:
   - TVL and TVL trend (7d)
   - Volume and volume trend (7d)
   - Fee APR (actual, not projected)
   - Bin step and price range
3. Check for red flags:
   - TVL < $50k
   - Declining volume
   - Single-sided liquidity
   - Whale concentration
4. Provide recommendation:
   - Entry: Yes/No/Wait
   - Suggested position size
   - Recommended bin range
   - Risk level (Low/Medium/High)

## Output Format

## Pool Analysis: {pool_name}

### Metrics
- TVL: $X (↑/↓ Y% 7d)
- Volume 24h: $X
- Fee APR: X%
- Bin Step: X

### Risk Assessment
[Analysis...]

### Recommendation
[Entry recommendation with reasoning]

3. Use the Skill

bash
# In Claude Code
/analyze-pool 7qbRF...

Skill Types

Analysis Skills

Deep-dive on specific data:

markdown
# Token Analysis

Analyze a token for spot trading.

## Trigger

/analyze-token

## Instructions

1. Fetch token data from Jupiter
2. Check holder distribution
3. Analyze price history
4. Assess liquidity depth
5. Provide buy/pass recommendation

Workflow Skills

Multi-step processes:

markdown
# Morning Routine

Daily market prep workflow.

## Trigger

/morning

## Instructions

1. Fetch overnight price changes
2. Check funding rates
3. Review open positions
4. Identify top opportunities
5. Generate daily brief

Alert Skills

Monitoring and notifications:

markdown
# Price Alert

Set a price alert.

## Trigger

/alert

## Arguments

- `asset` - Asset to monitor
- `price` - Target price
- `direction` - above/below

## Instructions

1. Confirm alert parameters
2. Add to monitoring list
3. Send Telegram when triggered

Research Skills

Web search and analysis:

markdown
# Market Research

Research a market topic.

## Trigger

/research

## Arguments

- `topic` - Topic to research

## Instructions

1. Web search for recent news
2. Find relevant data sources
3. Synthesize findings
4. Provide actionable summary

Best Practices

Be Specific

markdown
# Good
When calculating APR, use actual fee data from the last 7 days,
not projected or annualized estimates.

# Bad
Calculate the APR.

Include Examples

markdown
## Examples

**User:** /analyze-pool 7qbRF...

**Response:**
## Pool Analysis: SOL-USDC

### Metrics
- TVL: $2.4M (↑12% 7d)
- Volume 24h: $850k
...

Define Output Format

markdown
## Output Format

Always structure the response as:
1. Summary (1-2 sentences)
2. Key Metrics (bullet points)
3. Analysis (2-3 paragraphs)
4. Recommendation (action + reasoning)

Handle Edge Cases

markdown
## Edge Cases

If the pool doesn't exist:
- Return "Pool not found: {address}"

If data is stale (>1 hour old):
- Warn user about data freshness
- Proceed with available data

Advanced Features

Tool Integration

Reference available tools:

markdown
## Instructions

Use these tools:
- `fetch_pools` - Get pool data
- `get_positions` - Check existing positions
- `web_search` - Research market context

Conditional Logic

markdown
## Instructions

If TVL > $1M:
  - Full analysis with all metrics
  - Historical comparison

If TVL < $1M:
  - Basic analysis only
  - Add low-liquidity warning

Multi-Domain Skills

markdown
## Instructions

1. Check DLMM positions for the token
2. Check Perps funding rates
3. Check Spot holdings
4. Aggregate exposure across domains

Sharing Skills

Export

bash
# Copy skill to share
cp .claude/skills/user/my-skill.md ~/Desktop/my-skill.md

Import

bash
# Add community skill
cp ~/Downloads/cool-skill.md .claude/skills/user/

Community Skills

Share your skills with the community via GitHub:

  1. Fork the claudefi repo
  2. Add your skill to skills/your-skill-name/
  3. Follow the SKILL.md format
  4. Open a pull request

See skills/CONTRIBUTING.md for guidelines.

Monetize your trading strategies:

FeatureDescription
Premium SkillsSell your best strategies to other traders
Revenue SharingEarn from skill purchases and usage
Creator DashboardTrack downloads, earnings, and feedback

Interested in being an early creator? Open an issue on GitHub to discuss.

Skills Marketplace

Coming soon: Install skills directly from the marketplace.

bash
# Future feature
claudefi skills install @author/skill-name

Debugging Skills

Test Invocation

bash
# Test in Claude Code
/my-skill test-argument

Check Loading

bash
# List loaded skills
ls -la .claude/skills/user/

View Logs

Skills usage is logged in the database for debugging.

autonomous claude agents across defi