メニュー
Plan and coordinate parallel subagent lanes using OpenCaw roles, durable task artifacts, and tool-specific execution guidance.
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
SOC 職業分類に基づく
Plan original trading card game and card-battler visual direction for TCG/CCG card frames, card illustrations, board styling, tokens/minions, hand/deck/graveyard zones, VFX, and Hearthstone-like fantasy tabletop polish while keeping references IP-safe.
Plan and specify production-ready game art sheets, including environment tilesets, biome/terrain sets, autotile sets, prop atlases, and directional character animation sheets with 4, 6, 8, 16, or custom facing counts. Use when art roles need to create tilesets or sprite animation sets before implementation handoff.
Generate the host repository STYLE.md by composing one or more selected templates from ./.styles/.
Maintain, generate, or validate the host repository STYLE.md art contract from OpenCaw .styles templates. Use when visual work must preserve a project art direction, when a user asks to choose or change art styles, or when art roles need to verify that STYLE.md still references valid style templates.
Prepare game art assets for implementation handoff with style references, source/runtime file paths, atlas metadata, pivots, anchors, animation rows, collision assumptions, sorting layers, licensing, and validation checks. Use when art roles deliver assets to engineering, QA, game design, or documentation.
Review isometric and 2.5D game art for projection, tile metrics, anchors, elevation, depth sorting, occlusion, lighting, walkability, and gameplay readability. Use for isometric sprites, tiles, props, environments, VFX, maps, atlas handoffs, or style reviews.
| name | orchestrate-subagents |
| description | Plan and coordinate parallel subagent lanes using OpenCaw roles, durable task artifacts, and tool-specific execution guidance. |
Use when a user specifies an agent, developer, worker, or parallel execution count, or when computer-science/project-manager identifies work that can safely run in parallel.
Use before spawning subagents or delegating work so each lane has a resolved role, explicit ownership, dependencies, expected output, and verification evidence.
../.ai/tasks/<task_name>/SUBAGENTS.md lane plan../commands/resolve-role.sh before assigning work.explorer for read-only investigation lanes and worker for implementation lanes when Codex subagents are available.You are working as `<role-id>` on lane `<lane-id>`.
You are not alone in the codebase. Other agents or the main agent may be working in parallel. Do not revert edits made by others. Keep your work inside your assigned scope and adjust to compatible changes if you encounter them.
Scope:
- <files, module, or responsibility>
Write set:
- <paths this lane may edit, or none for read-only lanes>
Dependencies:
- <none or named blockers>
Expected output:
- <deliverable>
Verification:
- <commands, checks, artifacts, or review evidence>
Final response requirements:
- summarize what you did
- list files changed
- provide verification evidence
- call out blockers, conflicts, or residual risks
../commands/create-subagent-plan.sh "<task_name>" ["agent_count"] [--dry-run]
../commands/validate-subagent-plan.sh "<task_name|path>"
../commands/record-subagent-result.sh "<task_name>" "<lane_id>" "" "<summary_file>" [--dry-run]