Adaptive Learning Skill

Activation

This skill auto-activates after code reviews to suggest capturing learnings.

Triggers:

•/validation:code-review completes (pass or fail)
•/validation:code-review-fix completes
•Manual invocation via /validation:learn

IMPORTANT: This is a SKILL that suggests learning, not enforces it. Learning analysis is optional but recommended.

Enforcement

🧠 ADAPTIVE LEARNING PROMPTS:

After Code Review Completes

When code review completes:

code

✅ Code review complete.

**Capture learnings?**
Run `/validation:learn` to:
- Extract insights from this review
- Identify recurring patterns
- Update learning metrics
- Generate improvement suggestions

This helps the framework get smarter over time.

When Issues Were Found

If code review found issues:

code

📊 Code review found {N} issues ({critical}, {high}, {medium}, {low}).

**Learning opportunity detected.**
Run `/validation:learn` to analyze if these issues are recurring
and generate improvement suggestions to prevent them in future.

When Review Was Clean

If code review found no issues:

code

✨ Clean code review! No issues found.

**Good patterns identified.**
Run `/validation:learn` to capture what went well
and reinforce good practices in the framework.

Behavior

When this skill activates:

1. Check Learning Metrics Status

Read .claude/agents/learning/learning-metrics.md to check:

•When was last analysis run?
•How many reviews since last analysis?
•Are there pending suggestions?

2. Generate Appropriate Prompt

Based on review outcome and metrics status:

If issues found + no recent learning:

•Strong suggestion to run learning analysis

If clean review + no recent learning:

•Gentle suggestion to capture good patterns

If analysis run recently:

•Light reminder that learning is available

3. Provide Learning Summary (if metrics exist)

Show brief learning status:

code

**Current Learning Status:**
- Reviews analyzed: {N}
- Recurring issues: {N} pending attention
- Suggestions: {N} pending approval

Run `/validation:learning-status` for full dashboard.

Examples

✅ GOOD: Capturing Learnings

code

User: /validation:code-review

[Code review runs and completes]

Skill prompts:
"✅ Code review complete. Found 2 medium issues.

 Capture learnings? Run `/validation:learn` to analyze
 if these issues are recurring and suggest improvements."

User: /validation:learn

[Learning analysis runs, creates insights artifact]

"Learning analysis complete. 1 recurring issue identified:
 'Missing input validation on API endpoints (3 occurrences)'

 Run `/validation:suggest-improvement rule "Add API input validation pattern"`
 to create improvement suggestion."

✅ GOOD: Building Learning Over Time

code

Review 1: Found security issue → Learning captures it
Review 2: Same security issue → Learning flags as recurring
Review 3: Suggestion created → Rule updated with security pattern
Review 4: No security issues → Learning notes improvement!

Integration Points

With Code Review

After /validation:code-review completes:

•Skill activates
•Suggests learning analysis
•User can accept or skip

With Code Review Fix

After /validation:code-review-fix completes:

•Skill activates
•Suggests learning analysis (what issues were fixed?)

With System Review

Complementary, not overlapping:

•System Review = PIV process quality
•Learning = Technical patterns and rules

Checklist

Before skipping learning analysis, consider:

• Are there issues that might be recurring?
• Would capturing this pattern help future features?
• Has it been a while since last learning analysis?
• Are there pending suggestions to review?

Resources

See Also:

•.claude/commands/validation/learn.md - Learning analysis command
•.claude/commands/validation/suggest-improvement.md - Generate improvements
•.claude/commands/validation/learning-status.md - View metrics dashboard
•.claude/agents/learning/ - Learning artifacts directory