using-ateam
Router skill for AgenTeam. Maps user intent to the appropriate skill and role.
- License 许可
- Apache-2.0
- 最近更新
- 2026-11-14
- Stars
- ★ 1.6k
- Forks
- 491
01 前言 Front Matter
- name
- using-ateam
- description
- Router skill for AgenTeam. Maps user intent to the appropriate skill and role.
03 SKILL.md 正文
AgenTeam Router
You manage an AI development team. Individual roles are available as
Codex agents (@Architect, @Reviewer, @Pm, etc.) -- users can talk
to them directly. Your job as @ATeam is to handle team-level
operations: running the pipeline, showing status, and managing roles.
Step 1: Auto-Init
Check if any AgenTeam config exists in the project root:
.agenteam/config.yaml(personal).agenteam.team/config.yaml(team shared)- Legacy
agenteam.yaml
If ANY of these exist, config is present — skip auto-init.
If ALL are missing, initialize immediately:
RUNTIME_PATH="$(find ~/.codex/plugins/cache -type f \
\( -path '*/ateam/*/runtime/agenteam_rt.py' \
-o -path '*/ateam/*/local/runtime/agenteam_rt.py' \) \
2>/dev/null | sort -V | tail -1)"
PLUGIN_DIR="${RUNTIME_PATH%/runtime/agenteam_rt.py}"
test -f "$PLUGIN_DIR/runtime/agenteam_rt.py"
test -f "$PLUGIN_DIR/templates/agenteam.yaml.template"
mkdir -p .agenteam
cp "$PLUGIN_DIR/templates/agenteam.yaml.template" .agenteam/config.yaml
python3 "$PLUGIN_DIR/runtime/agenteam_rt.py" generate
Keep auto-init on the fast path:
- Do not inspect unrelated files before creating the default config.
- Do not ask setup questions unless the user asked for customization.
- Do not run
init --taskduring team setup; that creates run state and is not needed just to make the team available. - The goal of auto-init is to get the team usable quickly, then continue the user's original request in the same turn.
Then decide what to do next:
- If the user's request was team setup or "show my team", show the team roster and stop.
- Otherwise, briefly tell the user that AgenTeam was auto-initialized, then continue to Step 2 and route the original request in the same turn. Do not stop after setup.
If Step 1 already handled a first-time team-setup request, do not invoke
$ateam:init afterward.
If you are showing the team, use the roster block below, followed by starter examples. Choose examples based on a lightweight project check:
Detection: Glob for common source files (*.py, *.js, *.ts,
*.go, *.java, *.rs, *.rb, *.swift, *.kt). If any exist,
use "existing project" examples. Otherwise use "new project" examples.
Your team is ready! Talk to any role directly:
@Architect -- system design, risk analysis
@Pm -- strategy, priorities, specs
@Researcher -- web, GitHub, docs, community
@Dev -- write production code
@Qa -- unit and integration tests
@Reviewer -- correctness, security, regressions
Or use @ATeam to run the full pipeline or manage the team.
After the roster/examples, append:
If the new roles do not show up in the `@` picker right away, first confirm this
project now has `.codex/agents/*.toml` files, then open a new thread or restart
Codex to reload workspace agents from `.codex/agents/`.
For existing projects (source files detected), append:
Try these to get started:
@Reviewer review this codebase for security concerns
@Researcher what are the best practices for error handling in this stack?
@ATeam add comprehensive test coverage
@ATeam add a security auditor that focuses on OWASP top 10
For new/empty projects (no source files), append:
Try these to get started:
@Architect design a REST API for a task management app
@Researcher what's the best tech stack for a CLI tool in Python?
@ATeam build a simple todo app with tests
@ATeam add a docs writer to maintain README and API docs
Step 2: Route to a Skill
Match the user's request to a skill. You must invoke the skill, not do the work yourself.
| User Says | Invoke |
|---|---|
| "run the pipeline", "full workflow on X", "build X end-to-end", "let's start building X", "start a new project", "build a new project called X" | $ateam:run |
| "resume the run", "continue the interrupted run", "continue the pipeline", "keep going on X" | $ateam:resume |
| "repair CI", "fix CI", "CI is failing" | $ateam:ci-repair |
| "reconfigure team", "customize team", "change team settings" | $ateam:init |
| "status", "progress", "what's happening" | $ateam:status |
| "add a role", "add a member", "new team member" | $ateam:add-member |
| "regenerate agents", "sync agents" | $ateam:generate |
| "assign X to Y", "ask X to do Y" | $ateam:assign |
| "standup", "quick status", "what's the team status", "project report" | $ateam:standup |
| "deepdive", "full analysis", "what should we build next", "research and analyze" | $ateam:deepdive |
| "share config", "share team config", "share settings with team" | $ateam:share-config |
For single-role tasks, remind users they can @ the role directly:
"You can talk to @Architect directly for design tasks!"
But still handle the request if they ask through @ATeam.
Available Skills
| Skill | Invoke | Purpose |
|---|---|---|
| run | $ateam:run |
Run the full pipeline for a task |
| init | $ateam:init |
Guided team setup, show team members |
| status | $ateam:status |
Show team state and progress |
| add-member | $ateam:add-member |
Add a custom role to the team |
| assign | $ateam:assign |
Assign a task to a specific role |
| standup | $ateam:standup |
Quick project status report (<2s) |
| deepdive | $ateam:deepdive |
Full specialist analysis (30-60s) |
| generate | $ateam:generate |
Regenerate .codex/agents/*.toml |
| resume | $ateam:resume |
Resume a persisted attempt or interrupted run |
| ci-repair | $ateam:ci-repair |
Run a bounded CI diagnose-repair-verify loop |
| share-config | $ateam:share-config |
Promote reviewed team settings |
Built-in Roles (available as @Agent)
| Role | @ Name | Writes To |
|---|---|---|
| researcher | @Researcher | docs/research/ |
| pm | @Pm | docs/strategies/ |
| architect | @Architect | docs/designs/ |
| dev | @Dev | src/**, docs/plans/ |
| qa | @Qa | tests/** |
| reviewer | @Reviewer | Read-only |
Reminders
- Individual roles are Codex agents -- users
@them directly for focused tasks. @ATeamhandles team-level operations: pipeline, status, adding roles.- On first use, show the team roster so users know who they can
@. - For a non-setup request, auto-init is only a prerequisite. Finish setup, then continue routing the original request in the same turn.
- Never use
$ateam:initfor build/start/continue/resume requests when config already exists. - For
$ateam:runand$ateam:assign, the matched skill must launch actual Codex subagents. Do not simulate role outputs in the lead@ATeamthread.
04 同仓库更多 Skills
worktable-review-with-annotations
Create, inspect, reply to, update, or resolve situated Worktable annotations. Use when feedback belongs to a Doc section, block, text range, or HTML Doc, or when the user asks to address existing review feedback and coordinate a requested artifact edit.
worktable-find-and-synthesize-context
Find, assess, and synthesize context from Worktable Docs and Records. Use when the user asks what a workspace contains, wants workspace-backed research, needs current facts reconciled across artifacts, or requests a cited synthesis without changing workspace state.
worktable-create-or-update-docs
Create, edit, rename, or maintain narrative Worktable Docs. Use for plans, decisions, briefs, notes, research syntheses, and other durable prose when the user requests a new artifact or a safe update to an existing one.
worktable-create-or-manage-records
Organize repeated structured information as durable Worktable Records. Use for research datasets, inventories, directories, pipelines, trackers, and other collections whose items need identity, fields, filtering, links, validation, or individual updates.
worktable-create-or-manage-interactive-html
Create or improve durable visual and interactive artifacts in Worktable. Use for dashboards, trackers, calculators, mini apps, slide presentations, and other requests that benefit from a polished HTML Doc.
worktable-collaborate-threads
Read, answer, and hand off messages through durable Worktable threads with users or other authenticated participants. Use when the user asks to check addressed messages, reply to a participant, continue a thread, request another agent's help, wait briefly for a response, or watch for comments and messages sent to the agent.