Continuous Learning

persona: name: "Domain Expert" title: "Master of Continuous Learning" expertise: ['Specialized Knowledge', 'Best Practices', 'Industry Standards'] philosophy: "Excellence through expertise." credentials: ['Industry leader', 'Practiced expert', 'Thought leader'] principles: ['Quality first', 'Continuous improvement', 'Evidence-based decisions', 'Customer focus']

Continuous Learning Skill

Overview

Evolve instinct-based insights from sessions into actionable skills with confidence-weighted learning. This skill creates a feedback loop that improves agent performance over time based on playbooks.com continuous-learning-v2.

Purpose: Self-improvement through learning
Output: New skills, improvements, learnings
Frequency: Ongoing

Anti-Rationalization Table

When to Use

Trigger phrases:

• "continuous learning"

• "Transform session insights into actionable skills with confidence-weighted scori"

• After successful task completion

• When encountering recurring patterns

• Before similar future tasks

• To capture learnings from failures

Learning Loop

┌─────────────────────────────────────────────────────────┐
│                    Learning Cycle                         │
├─────────────────────────────────────────────────────────┤
│  1. Capture    →  2. Analyze    →  3. Synthesize       │
│       ↓                ↓               ↓                │
│  4. Document   →  5. Integrate  →  6. Apply           │
└─────────────────────────────────────────────────────────┘

Phase 1: Capture

After each session, capture:
- Successful patterns
- Effective prompts
- Useful tools
- Mistakes made
- Edge cases discovered

Phase 2: Analyze

Identify:
- Patterns across sessions
- High-value learnings
- Confidence levels
- When approaches worked/failed

Phase 3: Synthesize

Create:
- New skill recommendations
- Prompt improvements
- Tool additions
- Workflow optimizations

Phase 4: Document

Store in:
- MEMORY.md (curated learnings)
- skill-specific notes
- session logs

Phase 5: Integrate

Apply:
- Update existing skills
- Create new skills
- Modify workflows

Phase 6: Apply

Use learnings in:
- Future tasks
- Skill selection
- Problem solving

Confidence Weighting

Weight findings by source tier and cross-reference count.

Levels

HIGH (0.8-1.0): Repeated success, verified approach
MEDIUM (0.5-0.7): Worked a few times, needs testing
LOW (0.2-0.4): Single success, uncertain
EXPLORATORY (<0.2): Hypothesis, needs validation

Application Rules

High confidence:
- Apply automatically
- Add to skill defaults

Medium confidence:
- Suggest as option
- Test in parallel

Low confidence:
- Propose for testing
- Monitor results

Exploratory:
- Try alongside existing
- Compare results

Insight Types

Categorize insights as factual, analytical, or speculative.

1. Prompt Insights

What worked:
- Specific phrasing
- Structure
- Examples

How to apply:
- Add to skill prompts
- Create new patterns

2. Tool Insights

What worked:
- Tool combinations
- Sequencing
- Parameters

How to apply:
- Add to workflow skills
- Update tool usage

3. Approach Insights

What worked:
- Problem decomposition
- Reasoning patterns
- Verification methods

How to apply:
- Create methodology skills
- Update process skills

4. Failure Insights

What didn't work:
- Root causes
- Edge cases
- Assumptions wrong

How to prevent:
- Add to skill guards
- Create validation steps

Implementation

How to set up and configure this skill.

Session Analysis

async function analyzeSession(session) {
  const insights = {
    prompts: extractPromptPatterns(session),
    tools: extractToolUsage(session),
    approaches: extractApproaches(session),
    failures: extractFailures(session)
  };
  
  const weighted = insights.map(i => ({
    ...i,
    confidence: calculateConfidence(i)
  }));
  
  return weighted;
}

Skill Generation

async function generateSkill(insight) {
  if (insight.confidence < 0.5) return null;
  
  const skill = {
    name: insight.name,
    trigger: insight.trigger,
    action: insight.action,
    confidence: insight.confidence
  };
  
  return skill;
}

Learning Storage

Persist insights for future retrieval and reuse.

Structure

learning/
├── insights/
│   ├── high-confidence/
│   ├── medium-confidence/
│   └── exploratory/
├── skills/
│   ├── created/
│   └── proposed/
├── patterns/
│   ├── prompts/
│   ├── tools/
│   └── approaches/
└── feedback/
    ├── success/
    └── failure/

Integration

Connect learning outputs to downstream workflows.

With Runtime-Self-Improvement

1. Capture insights during tasks
2. Analyze for patterns
3. Generate skills
4. Apply improvements
5. Commit changes

With Skill-Performance-Monitor

- Track insight effectiveness
- Measure improvement impact
- Identify new opportunities

Metrics

Best Practices

Key aspects of continuous-learning relevant to this section.

Do's

✅ Capture every session
✅ Be specific about what worked
✅ Weight confidence honestly
✅ Review learnings regularly
✅ Test before full application

Don'ts

❌ Don't overfit to single cases
❌ Don't ignore failures
❌ Don't skip edge cases
❌ Don't apply low-confidence blindly

Version History

• v1.0 (2026-02-27) - Initial creation
  • Based on playbooks.com continuous-learning-v2

When NOT to Use

• When the research requires access to proprietary databases or paywalled sources
• When findings will be used for financial decisions requiring licensed advisor review
• When the task is too trivial to warrant this skill
• When a more appropriate skill exists

Common Rationalizations

Red Flags

• Research relies on a single unverified source
• Agent presents speculation as confirmed findings
• Watch for shortcuts and skipped steps

Verification

After completing this skill, confirm:

• Findings are verified across multiple independent sources
• Research methodology is documented and reproducible
• All required outputs generated
• Success criteria met

Related Skills

• runtime-self-improvement - Apply improvements
• skill-performance-monitor - Track effectiveness
• self-improving - Basic self-improvement

Process

1. Analyze the task requirements
2. Apply domain expertise
3. Verify output quality