Help us improve
Share bugs, ideas, or general feedback.
From smart-run
Two-phase orchestrator: Plan with Opus → Execute with Sonnet. Opus analyzes the task and produces a structured plan; Sonnet agents execute each step in parallel where possible. Use when asked to "smart run", "/smart-run", "플랜실행", or when the user wants Opus-quality planning with Sonnet-speed execution across multiple skills.
npx claudepluginhub intenet1001-commits/csncompany_2-0 --plugin cs-smart-runHow this skill is triggered — by the user, by Claude, or both
Slash command
/smart-run:smart-runThis skill is limited to the following tools:
The summary Claude sees in its skill listing — used to decide when to auto-load this skill
When invoked, you orchestrate **two distinct phases**:
Mandates invoking relevant skills via tools before any response in coding sessions. Covers access, priorities, and adaptations for Claude Code, Copilot CLI, Gemini CLI.
Share bugs, ideas, or general feedback.
When invoked, you orchestrate two distinct phases:
cmux 터미널($CMUX_SOCKET_PATH 설정됨)에서 실행 시 진행 상황을 사이드바에 표시한다:
# Phase 시작 시 호출
[ -n "$CMUX_SOCKET_PATH" ] && cmux set-status "smart-run" "running" --icon "gear"
[ -n "$CMUX_SOCKET_PATH" ] && cmux set-progress 0.0 --label "Smart Run 시작..."
각 Phase 전환 시:
cmux set-progress 0.2 --label "PLAN: Opus 분석 중..."cmux set-progress 0.5 --label "EXEC: Sonnet 실행 중..."cmux set-progress 1.0 --label "완료" + cmux notify --title "Smart Run 완료" --body "[태스크 요약]"Spawn ONE agent with model: "opus" to analyze the task and return a structured plan.
Prompt the Opus agent with:
You are a senior architect. The user wants: <task>
Produce a PLAN with these sections:
## Goal
One-sentence summary.
## Steps
Numbered list. Each step must specify:
- What to do
- Which skill or tool to use (if applicable)
- Input/output dependencies with other steps
- Whether it can run in PARALLEL with other steps (mark with [PARALLEL])
## Risks
Key risks or blockers to watch for.
## Definition of Done
How to verify the work is complete.
Be thorough but concise. This plan will be handed directly to execution agents.
Wait for the Opus agent to return the plan before proceeding.
Read the plan from Phase 1. For each step:
[PARALLEL]): spawn Sonnet agents simultaneously in a single messageSpawn each execution agent with model: "sonnet".
Prompt each Sonnet execution agent with:
You are an expert implementer. Execute this specific step from a larger plan:
## Your Step
<step description>
## Full Plan Context
<full plan from Opus>
## Prior Step Results (if any)
<results>
Execute completely. Return: what you did, files changed, and any output the next step needs.
After all execution agents complete, summarize:
When the user runs /smart-run <task> or /플랜실행 <task>: