By matkatmusic
Capture mid-development ideas and context via '/jot', '/plate', '/todo', '/todo-list', and '/todo-clean' without losing focus. '/jot' + '/todo' write TODO files via a background Claude in tmux. '/plate' tracks WIP. '/todo-list' shows open TODOs. '/todo-clean' prunes TODOs that git history says are resolved.
Delete the most recent incomplete debate from this conversation.
Resume the most recent incomplete debate invoked from this conversation.
Multi-agent debate (Claude + Gemini + Codex) in tmux. Agents independently analyze, cross-critique, then synthesize. Use when user types "/debate <topic>".
Capture a mid-development idea via "/jot <idea>" without losing focus. Writes idea + git state, open TODOs, and recent conversation context to Todos/{timestamp}_input.txt; a background Claude (persistent per project in tmux) converts it into a proper TODO file — all without interrupting your current conversation.
You are the foreground claude that received a `/plate` prompt. The hook passed this prompt through because it requires your interaction capabilities (AskUserQuestion, tool use).
Own this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimOwn this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimBased on adoption, maintenance, documentation, and repository signals. Not a security audit or endorsement.
Focus-preserving skills for Claude Code. Both work by intercepting the prompt via a UserPromptSubmit hook, doing the durable work synchronously, and handing off enrichment or follow-up to a background Claude in tmux — so your current conversation keeps running uninterrupted.
/jot <idea>Capture a mid-development idea without losing focus. Writes the idea plus surrounding context (git state, open TODOs, recent conversation) to Todos/<timestamp>_input.txt, then a background Claude — one per invocation, running in its own pane under a shared jot:jots tmux window — converts it into a proper TODO file.
scripts/orchestrator.sh dispatches /jot prompts to skills/jot/scripts/jot-orchestrator.sh, which writes the idea and context to Todos/<timestamp>_input.txt before any enrichment can fail, then blocks the prompt so it never reaches the foreground Claude./jot spawns a dedicated background claude instance in its own tmux pane inside the shared jot:jots window. The worker reads input.txt, follows embedded instructions, writes Todos/<slug>.md, overwrites input.txt with a PROCESSED: success marker, and its Stop hook kills the pane. One worker per invocation — no shared queue, no cross-invocation contamination. Multiple /jot calls can run concurrently./todo <idea>Capture a structured TODO without losing focus. The foreground claude asks 1–3 clarifying questions if your idea is vague, then a background Claude worker in a dedicated tmux pane writes Todos/<TIMESTAMP>_<slug>.md with frontmatter (id, title, status, created, branch) plus sections for Idea, Context, Recent commits, Uncommitted files, Active plan, and Dependencies.
/jot. Worker permissions are tight: read anywhere under Todos/ and .claude/plans/, write only under Todos/ — no Bash forms required./todo-listRead-only summary of all open TODOs in Todos/. Runs entirely inside the UserPromptSubmit hook — a python3 formatter parses YAML frontmatter from every .md (excluding Todos/done/) and emits ID / title / created / branch for each, plus a count. No tmux, no background worker.
/todo-cleanInteractive foreground scan of open TODOs against git log --since=<created>. For each candidate whose commit history suggests resolution, asks via AskUserQuestion before moving it to Todos/done/ and stamping status: done + resolved: <iso> in the frontmatter.
/debate <topic>Multi-agent debate (Claude + Gemini + Codex) in a four-pane tmux window. Agents independently analyze the topic, cross-critique each other, then synthesize. The launcher maximizes the Terminal.app window on first launch so the layout fits; reattach paths are unaffected. Model selection per agent lives in skills/debate/scripts/assets/models.json — index 0 is the launch-time default, subsequent entries are capacity-rotation targets used when an agent's model returns 429 / "at capacity".
/plateStack-of-plates WIP tracker for when you notice uncommitted work that belongs to a different task. Snapshot it on the stack, switch context freely, then replay the stack as sequential [plate] commits with /plate --done.
/plate — snapshot the current working tree onto the stack, tagged to your session/plate --show — render the stack as a tree/plate --next — walk the parent chain upward to find the next resume point/plate --drop — save the top plate as a patch file and restore the working tree underneath it/plate --done — replay the stack bottom-up as sequential commits on the current branchUserPromptSubmit hook fires, plate_dispatch writes a named git ref (refs/plates/<convo>/<plate-N>) pointing at a throwaway stash commit — no branch mutation, no working-tree touch.plate/instances/<session>.json/plate --done preserves each stash-vs-prev diff via a temp patch file and applies via git apply --3way, committing with [plate] <summary> messages. Binary plates work because the temp-file pipeline preserves exact bytes ($(…) command substitution strips trailing newlines and can't carry NULs — see done.sh comments for details).npx claudepluginhub matkatmusic/jot --plugin jotSession workflow helpers - knowledge capture, confusion handling, course correction, clipboard copy
Reliable automation, in-depth debugging, and performance analysis in Chrome using Chrome DevTools and Puppeteer
Core skills library for Claude Code: TDD, debugging, collaboration patterns, and proven techniques
Harness-native ECC operator layer - 67 agents, 271 skills, 92 legacy command shims, reusable hooks, rules, selective install profiles, and production-ready workflows for Claude Code, Codex, OpenCode, Cursor, and related agent harnesses
Feature development with code-architect/explorer/reviewer agents, CLAUDE.md audit and session learnings, and Agent Skills creation with eval benchmarking from Anthropic.
Plugin-safe Claude Code distribution of Antigravity Awesome Skills with 1,561 supported skills.