القائمة
Keep orchestrator chat lean — artifacts hold full agent output; orchestrator routes by file pointer only. Applies to all /kon:* commands that spawn Task subagents.
التثبيت باستخدام Codex أو Claude انسخ هذا Prompt والصقه في Codex أو Claude أو مساعد آخر ليراجع صفحة Skill ويثبّتها لك.
استنادا إلى تصنيف SOC المهني
Multi-agent design argument — Azusa challenges Mugi's plan, Mugi revises. Used by /kon:design after exploration and initial plan draft.
Multi-agent software engineering workflow. Use this skill whenever the user types a /kon: command — including /kon:team, /kon:design, /kon:quick, /kon:debug, /kon:ask, /kon:review, /kon:review-pr, /kon:hunt, /kon:gc, /kon:research, /kon:describe-issue, /kon:begin, /kon:finish, /kon:summarize, /kon:retro, /kon:address-comments, /kon:todo, /kon:understand-codebase. Orchestrates specialized agents: Azusa (explorer), Jun (researcher), Mugi (planner), Yui (implementer), Mio (reviewer), Sawako (cleaner), Nodoka (summarizer).
This skill should be used by all kon agents at startup, before beginning work, to load relevant memory from public and repo indexes with relevance-based ordering and a 10-entry cap. Consumers: Azusa, Mugi, Mio, Jun, and any future memory-reading agent.
Convention for per-repo skill directories at ~/.kon/projects/<repo-name>/skills/<name>/SKILL.md. Each named subdirectory is one skill. Agents load all SKILL.md files as the first step before memory entries.
When requirements, code behavior, root cause, or review scope is unclear — ask the user before proceeding. Never invent facts, assumptions, or fixes. Applies to every kon stage and every agent.
Route plain-text user messages while a /kon:begin session is open. Reuse the same session JSON — never init a second session for sub-turns.
| name | orchestrator-context |
| description | Keep orchestrator chat lean — artifacts hold full agent output; orchestrator routes by file pointer only. Applies to all /kon:* commands that spawn Task subagents. |
Owner: orchestrator
Consumers: all /kon:* commands that spawn Task subagents (team, design, quick, debug, ask, research, review, begin routing, …)
Core principles (always): follow skills/core-principles — route by file pointer; don't restate or amplify subagent output (simplicity in orchestrator context).
Subagent Task returns enter the orchestrator's context window. Do not amplify bloat by quoting, restating, or forwarding that output in chat or in later spawn/resume prompts.
Full output lives in artifacts. Orchestrator holds pointers only.
| Agent | Full output artifact | When |
|---|---|---|
| Orchestrator | sessions/<SID>/understanding.md | team, design, debug (orchestrator writes after pre-plan gate) |
| 🎸 Azusa | sessions/<SID>/explore.md | /kon:team, /kon:design (subagentStop hook) |
| 🎸 Azusa | sessions/<SID>/debug.md | /kon:debug (orchestrator writes once after Azusa) |
| 📚 Jun | .kon/research.md | external research steps |
| 🍰 Mugi | PLAN_FILE (.kon/plan-<SID>.md or session plan) | after plan step |
| 📝 Mio | sessions/<SID>/review.md | team, quick, review, debug (subagentStop hook) |
| 📝 Mio | sessions/<SID>/pr-review.md | /kon:review-pr |
| 📚 Jun | sessions/<SID>/issue-summary.md | /kon:describe-issue |
| 📋 Nodoka | sessions/<SID>/summary.md | /kon:summarize |
Resolve paths:
SID="<session-id>"
SESSION_DIR=$(python3 $KON_ROOT/scripts/kon_session.py session-dir --id "$SID")
REVIEW_FILE=$(python3 $KON_ROOT/scripts/kon_session.py artifact-path --id "$SID" --name review.md)
EXPLORE_FILE=$(python3 $KON_ROOT/scripts/kon_session.py artifact-path --id "$SID" --name explore.md)
review.md directly. See skills/failure-handling.## Decisions needed bullets; do not dump the full plan into chat (user opens PLAN_FILE).complete-agent with a one-sentence summary; do not copy the Task body into the summary field.Mugi (after pre-plan gate on team/design/debug):
PLAN_FILE: .kon/plan-<SID>.md
UNDERSTANDING_FILE: sessions/<SID>/understanding.md
EXPLORE_FILE: sessions/<SID>/explore.md
Read UNDERSTANDING_FILE, EXPLORE_FILE, and .kon/research.md (if present). Write the plan. No implementation.
Yui resume (after Mio block):
Resume. Read must-fix from sessions/<SID>/review.md — fix each by number.
PLAN_FILE: … Milestone: N. Report files changed only.
Mio resume:
Resume. Re-review milestone N. Prior must-fix: sessions/<SID>/review.md.
Verify each item with git diff evidence. Full hook-compliant output.
Every subagent ends with ## Orchestrator handoff (≤5 lines) so the orchestrator can route without re-reading the full body:
## Orchestrator handoff
- **Verdict**: APPROVED | BLOCKED | done | …
- **Artifact**: `sessions/<SID>/review.md` (or other path)
- **Next**: resume Yui | spawn Mugi | wait for user | …
- **Note**: one sentence only
Hooks persist the full output to artifacts; the handoff block is for orchestrator routing only.
/kon:begin)When a begin session routes into team/design/debug, the same rules apply.
For design → implement, prefer a new chat + /kon:team with existing PLAN_FILE to drop explore/plan debate from orchestrator context.
skills/teammate-flowskills/failure-handlingskills/session-tracking