team-reset
Transparent agent reset — shut down unresponsive agents and re-orient the implicit team, reporting roster changes
- License 许可
- Apache-2.0
- 最近更新
- 2026-09-21
- Stars
- ★ 0
- Forks
- 0
01 前言 Front Matter
- args
- team_name
- name
- team-reset
- description
- Transparent agent reset — shut down unresponsive agents and re-orient the implicit team, reporting roster changes
03 SKILL.md 正文
Handle the agent reset lifecycle for the {team_name} team.
Harness note (2026-06-16): the current Claude Code harness has no
TeamCreate/TeamDeletetools. The session runs on a single implicit team — there is nothing to delete or recreate. "Team reset" in this harness therefore means shutting down running agents and re-orienting, not tearing a team config down and rebuilding it. The steps below have been updated accordingly; the old TeamDelete/TeamCreate/config-file steps no longer apply.
Instructions
1. Report current team state
Read the current team roster from .claude/team/roster/ and report to the user:
**Current team: {team_name}**
| Role | Name | Status |
|------|------|--------|
| Manager | Fatima Okonkwo | Active |
| ... | ... | ... |
List all active roster members with their roles and status.
2. Send shutdown requests
Send a shutdown_request message via SendMessage to ALL agents you spawned this session:
{"type": "shutdown_request", "reason": "Team reset initiated"}
Send one message per agent (structured messages cannot be broadcast). Wait ~5–30 seconds for agents to acknowledge and terminate.
3. Handle unresponsive agents
There is no TeamDelete to force in this harness, so a stuck agent cannot be cleared by deleting a team config. If an agent does not acknowledge its shutdown request:
- Report to user: "{N} agent(s) unresponsive: {names}."
- Re-send the
shutdown_requestonce more. - If it still does not terminate, surface it to the user — a lingering agent is a UI/resource annoyance, not a blocker; spawning fresh agents is unaffected because the team is implicit. Do NOT attempt to edit or delete any
~/.claude/teams/...config (that mechanism is gone).
4. Re-orient (no recreate needed)
The implicit {team_name} team persists for the session — there is nothing to recreate. Re-read the roster files in .claude/team/roster/ so you have the current roster in context for the next round of spawns.
Report to user:
**Implicit team re-oriented:** {team_name}
| Role | Name | Status |
|------|------|--------|
| Manager | Fatima Okonkwo | Active |
| ... | ... | ... |
5. Highlight roster changes
If there are differences between the prior roster and the current roster files (hires, departures, role changes), explicitly report them:
**Roster changes:**
- DEPARTED: {name} ({role}) — {reason}
- HIRED: {name} ({role}) — replacing {departed name}
- ROLE CHANGE: {name} — {old role} → {new role}
If no changes: "Roster unchanged from previous round."
6. Ready confirmation
Confirm the team is ready for work:
Team `{team_name}` is ready ({N} roster members). Spawn agents via the Agent tool
(team_name: {team_name}) when ready — the orchestrator is the sole spawner.
What remains manual
- The orchestrating Claude instance must still spawn individual agents via the
Agenttool (agents cannot self-spawn) - Roster file changes (hires/fires) must be committed separately
04 同仓库更多 Skills
wave-wrapup
Finalize a wave — PR review, merge sequencing, issue cleanup, worktree cleanup, and handoff to retro
wave-start
Initialize a new wave — worktree cleanup, branch creation, label setup, status file update
wave-scope
Reconcile declared-vs-labeled wave scope between /wave-retro and /wave-kickoff — collect dispositions, fold in retro carry-forwards and memory must-includes, and refresh the meta-issue
wave-retro
Automated wave retrospective — PR analysis, assessments, trust matrix updates, feedback log, charter change proposals
wave-kickoff
Automated wave planning — branch creation, label management, issue labeling, kickoff comments, and execution plan
wave-audit
Audit open issues against merged PRs — find and close orphaned issues with proper comments