System Upgrade Pipeline

Operator Context

This skill orchestrates systematic upgrades to the agent/skill/hook/script ecosystem when external changes warrant adaptation. It is a top-down upgrade mechanism — triggered by Claude Code releases, user goal changes, or accumulated retro learnings — complementing the bottom-up retro-knowledge-injector.

Hardcoded Behaviors (Always Apply)

• Show Plan Before Implementing: Phase 3 output (ranked upgrade list) MUST be presented to the user and approved before Phase 4 begins. Never silently execute upgrades.
• Reuse Domain Agents: Phase 4 (IMPLEMENT) dispatches to existing domain agents (skill-creator-engineer, agent-creator-engineer, hook-development-engineer, golang-general-engineer, etc.). The upgrade engineer orchestrates; specialists execute.
• Parallel Fan-Out: When 3+ components need the same type of upgrade, dispatch in parallel using multiple Agent tool calls in a single message.
• Score Delta Required: Phase 5 (VALIDATE) must produce before/after evaluation delta, not just "looks good." Use agent-evaluation skill.
• Trigger Type Determines Input: The three trigger types (claude-release, goal-change, retro-driven) require different input parsing in Phase 1.

Default Behaviors (ON unless disabled)

• Scope Limiting: Default audit depth = 10 most-recently-modified agents + all hooks. Full audit only if user says "comprehensive" or "all".
• Dry Run Presentation: Show Phase 3 output as a formatted table with Tier (critical/important/minor) and effort estimate.