Improve Skills with AI¶

Use AI to enhance, optimize, and fix existing skills.

Overview¶

Moltler can analyze your skills and suggest improvements:

IMPROVE SKILL my_skill;

Output:

Analyzing skill: my_skill v1.0.0

Suggestions:
  1. Add error handling for external API calls
  2. Implement retry logic for transient failures
  3. Add logging for debugging
  4. Cache results to reduce API calls

Apply suggestions? [Y/n]

Automatic Improvements¶

Apply All Suggestions¶

IMPROVE SKILL my_skill
APPLY ALL;

Apply Specific Suggestions¶

IMPROVE SKILL my_skill
APPLY [1, 3];  -- Apply suggestions 1 and 3

Preview Changes¶

IMPROVE SKILL my_skill
PREVIEW;

Shows diff:

  BEGIN
+     TRY
          DECLARE result = HTTP_GET(url);
+         AGENT_LOG('API call succeeded', data => result);
-         RETURN result;
+         RETURN result;
+     CATCH
+         AGENT_LOG('API call failed', level => 'error', error => ERROR_MESSAGE());
+         RAISE;
+     END TRY;
  END SKILL;

Improvement Types¶

Error Handling¶

IMPROVE SKILL my_skill
ADD error_handling;

Adds try/catch blocks, graceful degradation, meaningful error messages.

Performance¶

IMPROVE SKILL my_skill
OPTIMIZE performance;

Adds caching, pagination, query optimization, parallel execution.

Logging¶

IMPROVE SKILL my_skill
ADD logging;

Adds structured logging for debugging and observability.

Tests¶

IMPROVE SKILL my_skill
ADD tests;

Generates comprehensive test cases.

Documentation¶

IMPROVE SKILL my_skill
ADD documentation;

Adds inline comments, parameter descriptions, usage examples.

Security¶

IMPROVE SKILL my_skill
HARDEN security;

Adds input validation, secret handling, rate limiting.

Skill Analysis¶

Analyze Without Changing¶

ANALYZE SKILL my_skill;

Output:

Skill Analysis: my_skill v1.0.0

Code Quality:
  ✓ Syntax valid
  ✓ No unused variables
  ⚠ Missing error handling (3 locations)
  ⚠ No logging statements

Performance:
  ✓ Query is efficient
  ⚠ Could benefit from caching
  ⚠ Consider pagination for large results

Security:
  ✓ No hardcoded secrets
  ⚠ Input not validated
  ⚠ Error messages may leak info

Test Coverage:
  ✗ No tests defined
  Recommended: 5 test cases

Documentation:
  ✓ Description provided
  ⚠ Parameters not documented

Compare Versions¶

COMPARE SKILL my_skill
VERSION '1.0.0' WITH '2.0.0';

Fix Issues¶

Auto-Fix¶

FIX SKILL my_skill;

Automatically fixes:

Syntax errors
Type mismatches
Undefined variables
Missing returns

Fix Specific Issue¶

FIX SKILL my_skill
ISSUE 'timeout_handling';

Fix Based on Errors¶

-- Skill failed with error
FIX SKILL my_skill
BASED ON ERROR 'Connection timeout after 30s';

AI suggests:

-- Add timeout handling
DECLARE result = HTTP_GET(url, timeout => 60);

-- Or add retry logic
DECLARE attempts = 0;
WHILE attempts < 3 LOOP
    TRY
        RETURN HTTP_GET(url);
    CATCH timeout_error THEN
        SET attempts = attempts + 1;
        WAIT INTERVAL '2s';
    END TRY;
END LOOP;

Refactoring¶

Extract Common Logic¶

REFACTOR SKILL my_skill
EXTRACT FUNCTION 'validate_input';

Split Skill¶

REFACTOR SKILL large_skill
SPLIT INTO [
    'fetch_data',
    'process_data',
    'store_results'
];

Merge Skills¶

REFACTOR SKILLS [skill_a, skill_b]
MERGE INTO 'combined_skill';

Learning from Feedback¶

Improve Based on Usage¶

IMPROVE SKILL my_skill
BASED ON USAGE;

Analyzes:

Common parameter values
Frequent errors
Performance patterns
User feedback

Improve Based on Outcomes¶

IMPROVE SKILL my_skill
BASED ON OUTCOMES
WHERE success_rate < 0.9;

Batch Improvements¶

Improve All Skills¶

IMPROVE ALL SKILLS
WHERE last_updated < NOW() - INTERVAL '90d';

Improve Skill Pack¶

IMPROVE SKILL PACK observability_tools;

What's Next?¶

Recommendations

Get proactive suggestions.

Recommendations