Skill UI

Captures user corrections during Claude sessions, queues them for review, and syncs accepted learnings into CLAUDE.md so future runs use the verified tools and conventions; run `/reflect` whenever you complete meaningful work or explicitly ask Claude to remember something.

Use when you have a written implementation plan to execute in a separate session with review checkpoints

Streamline post-meeting work by reviewing Granola notes, enhancing summaries, extracting action items, sharing across stakeholders, and archiving decisions so teams stay aligned after every session.

Simulate a structured peer-review workflow where a lead designer runs a brainstorming session and sequential specialist agents (skeptic, constraint guardian, user advocate, arbiter) validate assumptions, flag risks, and lock decisions before implementation.

Transforms single-line goals into multi-session, multi-agent engineering plans with dependency maps, parallel step detection, adversarial review, and mutation-safe change protocols so new agents can execute each step independently.

Orchestrates parallel AI-agent sessions with dmux tmux panes, letting you split research, implementation, testing, and reviews across Claude Code, Codex, OpenCode, and other tools while merging outputs back into a coordinated flow.