Menu
Task subagents must use the orchestrator's model — Cursor defaults to Composer if model is omitted. Applies whenever spawning Task subagents in kon.
Install with Codex or Claude Copy this prompt, paste it into Codex, Claude, or another assistant, and let it review the skill page and install it for you.
Based on SOC occupation classification
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 | model-inheritance |
| description | Task subagents must use the orchestrator's model — Cursor defaults to Composer if model is omitted. Applies whenever spawning Task subagents in kon. |
Owner: orchestrator
Consumers: all /kon:* commands that spawn Task subagents
Core principles (always): follow skills/core-principles — simplest correct spawn config; never omit model and hide the mismatch.
Cursor does not inherit the parent chat model on Task subagents. If you omit model, subagents often run on Composer while the orchestrator uses Opus, Sonnet, etc.
Per-agent model: lines in agents/*.md frontmatter are not applied to Task spawns — ignore them for model selection.
Every Task spawn and resume must pass model=<orchestrator slug>.
Use the same slug on resume passes (fix/re-review loops).
MODEL=$(python3 $KON_ROOT/scripts/kon_session.py get-orchestrator-model --id "$SID")
Your runtime — if the query above is empty, use the model slug from your Cursor session context (the model the user selected in the composer bar).
Overrides — KON_ORCHESTRATOR_MODEL env or ~/.kon/config.json → orchestrator_model.
Task(
subagent_type="generalPurpose",
model="<same slug as orchestrator>",
prompt="Read agents/Azusa.md …",
description="Azusa explore",
)
Do not rely on subagent_type="explore" alone — still pass model.
After spawning, the subagent label in Cursor should match the orchestrator family (e.g. Opus 4.8 ↔ claude-opus-4-8-thinking-high), not Composer 2.5.