Skill Regeneration (L2 - Directed)

You are executing the skill-regeneration skill.

User Request

$ARGUMENTS

Purpose

Refine and update existing skills based on concrete practices learned during execution. Keep skill definitions minimal and actionable.

Autonomy Level: L2 (Directed)

• Trigger: Manual only (human-initiated)
• Human: Provide practice insights, approve changes
• AI: Analyze → distill → update → validate

Agent Loop

while not goal_reached:
    researcher.read_target_skill()
    researcher.analyze_practice_log(concrete_examples)
    writer.distill_to_minimal_patterns()
    coder.update_skill_md(target_skill)
    human_checkpoint("Review skill update")

Input Required

• Target skill name to update
• Concrete practice examples or session logs
• (Optional) Specific sections to refine

Principles

1. Concise over comprehensive - Remove redundant explanations
2. Concrete over abstract - Real examples beat theory
3. Actionable over descriptive - Steps beat prose
4. Patterns over instances - Generalize from specifics

Output

Updated SKILL.md with:

• Tighter agent loop
• Refined checkpoints
• Distilled patterns from practice

Human Checkpoint

• Review diff of skill changes
• Confirm patterns match intent

Output

• PR to skills repo with updated SKILL.md