Use when manage multiple local CLI agents via tmux sessions (start/stop/monitor/assign) with cron-friendly scheduling.
日本語の概要は準備中です。原文の説明を表示しています。
Orchestrates the dedicated Kinect wall harp implementation team. Use whenever the user asks to build, implement, continue, fix, QA, document, or ship the Kinect wall harp, Kinect v2 wall-depth harp, FreenectTD depth-blob hand tracking, projected wall strings, pluck synth harp, or this feature's Layer 1 tool/recipe/docs. Use this before the generic tdmcp-pipeline for this feature.
インストールする前に、エージェントに与えられる指示の中身を確認できます。
Coordinate the dedicated team for the approved Kinect wall harp prototype. This skill specializes the generic tdmcp pipeline for a physical Kinect v2 + projector wall instrument where FreenectTD depth behavior, manual calibration, two-hand blob tracking, internal audio, and honest live validation all matter.
This repo documents that the current environment runs teams as coordinated
sub-agents rather than TeamCreate. Use a hybrid sub-agent workflow:
| Phase | Mode | Reason |
|---|---|---|
| Plan | local lead or kinect-wall-harp-lead | keep scope tied to the approved spec and wall setup. |
| Prototype + tool build | parallel sub-agents when write scopes are disjoint | live TD prototype and isolated tool file can advance separately. |
| Integrate/docs | single-writer sub-agent | registries, CLI, recipes, and docs are shared files. |
| QA/fix loop | sub-agent fan-in | QA routes precise defects to the owner; live hardware gets PASS / UNVERIFIED accounting. |
All spawned agents use the inherited model unless the user explicitly asks for a different one.
| Agent | Role | Skills | Output |
|---|---|---|---|
kinect-wall-harp-lead | scope owner and wave captain | this skill | _workspace/kinect-wall-harp/00_plan.md |
kinect-wall-harp-prototyper | live FreenectTD/Kinect wall prototype and calibration findings | tdmcp bridge/tools | _workspace/kinect-wall-harp/01_prototype.md |
kinect-wall-harp-tool-builder | new Layer 1 tool file + focused unit test | td-feature-build | _workspace/kinect-wall-harp/02_build_tool.md |
kinect-wall-harp-integrator | single-writer registry, CLI, recipe, docs | td-feature-integrate | _workspace/kinect-wall-harp/03_integrate.md |
kinect-wall-harp-qa | offline gates, boundary QA, live TD/Kinect validation | td-feature-qa | _workspace/kinect-wall-harp/04_qa.md |
Use existing tdmcp-bridge-engineer only if implementation proves that a new
bridge REST endpoint is necessary. The default plan should avoid a new endpoint
and use the existing bridge execution path.
Read these before changing code:
docs/superpowers/specs/2026-06-23-kinect-wall-harp-design.mdCLAUDE.mdFreenectTOP, Kinect v2 depth buffer at Render
Select index 1, point cloud index 2, IR index 3.src/tools/layer1/createBlobReactive.tssrc/tools/layer1/createMotionReactive.tssrc/tools/layer1/createProjectionMapping.tssrc/tools/layer1/ when present.git status --short._workspace/kinect-wall-harp/./api/info before any live prototype or QA
claim.FreenectTOP is registered if the bridge permits live probes.Create _workspace/kinect-wall-harp/00_plan.md with:
Default slice order:
Run in parallel when both are in scope and write scopes are disjoint:
kinect-wall-harp-prototyper: use the live bridge to build or validate
/project1/kinect_wall_harp, prove the depth-wall mask, and record
calibration/hand-channel findings. If bridge is offline, produce a live
checklist instead.kinect-wall-harp-tool-builder: create only the new tool file and focused
unit test. Do not edit shared files.Both agents must write their _workspace/kinect-wall-harp/0*_*.md notes.
Invoke kinect-wall-harp-integrator after the tool builder reports export
names. It owns all shared files:
src/tools/layer1/index.ts;src/cli/agent.ts;recipes/kinect_wall_harp.json if the recipe slice is active;The integrator must run npm run typecheck and npm run build. If docs/recipes
changed, also run npm run validate:recipes and the docs generation/build gate
that matches the touched files.
Invoke kinect-wall-harp-qa incrementally:
QA sends precise defects to the owner and re-validates after fixes. Cap repeated fix loops at 2-3 rounds, then report the blocker.
Report:
.tox, MIDI/OSC, or production audio.Do not tag, release, or push unless the user explicitly asks.
approved spec
-> lead plan
-> live prototype notes -----+
-> isolated tool builder ----+-> integrator -> qa -> fixes -> final report
| Situation | Strategy |
|---|---|
| Bridge offline | Build/test offline; mark live checks UNVERIFIED pending bridge. |
| FreenectTD unavailable | Keep synthetic fallback green; report plugin install/load state. |
| Kinect connected but depth not cooking | Stop live claims; surface raw node errors and next hardware checks. |
| Two blobs merge | Treat as one blob and document limitation; do not claim two-hand pass. |
| Shared-file conflict | Stop shared edits and ask the lead to reconcile. |
| QA fail after 3 rounds | Hold failing slice; report evidence and next fix. |
User says "implementar a harpa Kinect." The orchestrator reads the approved
spec, creates _workspace/kinect-wall-harp/00_plan.md, runs prototype and tool
build slices, integrates the tool and CLI, runs typecheck/build/focused tests,
validates synthetic TD output if the bridge is up, and reports live Kinect wall
tracking as PASS or UNVERIFIED.
The bridge is up but FreenectTOP is not registered. The orchestrator still
builds the synthetic-safe tool and docs, QA marks FreenectTD/Kinect checks
UNVERIFIED, and the final report names the exact plugin/load step to rerun.
Should trigger:
Should not trigger:
まだレビューはありません。使ってみた感想をお寄せください。
概要と使いどころ
Use when manage multiple local CLI agents via tmux sessions (start/stop/monitor/assign) with cron-friendly scheduling.
日本語の概要は準備中です。原文の説明を表示しています。
Use when a hybrid memory system that provides persistent, searchable knowledge management for AI agents (Architecture, Patterns, Decisions).
日本語の概要は準備中です。原文の説明を表示しています。
Reference for how BDB structures autonomous software engineering work — the seven-node dispatcher graph (Architect, TechLead, UI/UX, Engineering, Media/EventTech, Reviewer, Shipping) that /startcycle-graph actually runs. Use when you need the high-level lifecycle framing without inventing your own process.
日本語の概要は準備中です。原文の説明を表示しています。
Tools are how AI agents interact with the world. A well-designed tool is the difference between an agent that works and one that hallucinates, fails silently, or costs 10x more tokens than necessary. This skill covers tool design from schema to error handling.
日本語の概要は準備中です。原文の説明を表示しています。
Harness patterns for coding agents — memory, permissions, context engineering, delegation, skills, hooks, bootstrap.
日本語の概要は準備中です。原文の説明を表示しています。
agenttrail: live map of a multi-agent build in the browser: which plan component is being worked on, by which agent or harness, what is done and what is stuck. Use when a multi-agent pipeline starts (/startcycle, /startcycle-graph, /teamwork-preview) or after a plan-canvas approve, when the user asks to see what the agents are doing ("live map", "build board", "who is running now"), or by yourself whenever a multi-agent run is under way — no user prompt needed.
日本語の概要は準備中です。原文の説明を表示しています。