What problem does it solve?
Non-trivial code changes often fail because they start without a grounded plan: dependencies get ordered wrong, risky migrations lack rollback paths, and completion criteria go unverified. This Skill turns vague feature requests into concrete implementation maps rooted in actual repository evidence.
Core Features & Use Cases
- Evidence-First Planning: Reads AGENTS.md, architecture docs, current diffs, and critical code before proposing anything, so plans never invent files, APIs, or migrations.
- Structured Plan Contents: Each step names outcomes, critical files, dependency ordering, data/compatibility/concurrency risks, and direct verification signals.
- Mode-Aware Execution: Stays read-only in Plan Mode, but continues implementation automatically outside it when the next step is clear, and coordinates with goal tracking and multi-agent execution skills.
- Use Case: Ask it to plan a cross-crate feature touching Rust and TypeScript code; it traces the owning code paths, orders the dependent steps, flags migration risks, defines verification for each criterion, then proceeds to implement.
Quick Start
Ask the agent to plan this cross-crate feature with rollback and verification steps, then implement it.
