メニュー
Overview of how to coordinate the BWOC agent fleet from OpenAI Codex. Explains when to use each bwoc verb (list/status/send/run/chat/task/team/memory) and how they fit together. Start here for any multi-agent coordination request.
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
SOC 職業分類に基づく
Open an interactive chat session with a BWOC agent using its manifest-driven backend/model. Wraps `bwoc chat`. Use for back-and-forth conversation rather than one-shot tasks.
Read and search BWOC workspace-level memory (`.bwoc/memory/`) and Tier-2 deep memory. Wraps `bwoc memory`. Use to recall prior decisions, notes, and session context.
Run a single task on a BWOC agent non-interactively and capture the result (headless). Wraps `bwoc run`. Use when you need a one-shot answer/output back from an agent.
Append a message to a BWOC agent's inbox (fire-and-forget). Wraps `bwoc send`. Use to hand work or notes to an agent asynchronously.
Show a per-agent health and identity snapshot for the BWOC fleet. Wraps `bwoc status`. Read-only — use to inspect a single agent or summarize all agents.
Manage a BWOC team's shared task list — add, list, claim, complete, plan, approve, reject. Wraps `bwoc task`. Use to coordinate work across a team.
| name | bwoc-fleet |
| description | Overview of how to coordinate the BWOC agent fleet from OpenAI Codex. Explains when to use each bwoc verb (list/status/send/run/chat/task/team/memory) and how they fit together. Start here for any multi-agent coordination request. |
This plugin exposes the BWOC agent fleet into OpenAI Codex by wrapping the bwoc CLI. Every skill is a thin shell-out — there is no server or daemon. The BWOC workspace (with .bwoc/agents.toml) must be reachable from where Codex runs, and bwoc must be on PATH.
Codex ──@skill──▶ skill instructions ──exec──▶ bwoc CLI ──▶ BWOC workspace
(agents, teams,
tasks, memory)
Agents are persistent, named, backend-neutral workers (Claude, Codex, Antigravity, Kimi, Copilot, Ollama, …). Teams are named subsets sharing a task list. Memory is workspace-level recall.
| Goal | Skill | Command |
|---|---|---|
| Discover what agents exist | bwoc-list | bwoc list |
| Inspect one agent's health/identity | bwoc-status | bwoc status <name> |
| Hand work to an agent, async (no reply) | bwoc-send | bwoc send <to> "<msg>" |
| Get a one-shot result back (headless) | bwoc-run | bwoc run <agent> --task "..." |
| Interactive back-and-forth | bwoc-chat | bwoc chat <name> |
| Coordinate work across a team | bwoc-task | bwoc task <add|list|claim|complete|plan|approve> |
| Organize agents into groups | bwoc-team | bwoc team <create|list|retire> |
| Recall prior decisions/notes | bwoc-memory | bwoc memory <list|show|search|wake-up|t2-search> |
bwoc list (optionally --status active --json) to see who's available and which backend they run.bwoc status <name> on candidates to confirm health/identity.bwoc run <agent> --task "..." and relay the captured output.bwoc send <agent> "..." (fire-and-forget).bwoc team create <id> --members ..., then drive the shared list with bwoc task add/list/claim/complete. Plan-gated tasks use task plan → task approve → task complete (Pavāraṇā).bwoc memory search "<topic>" or bwoc memory wake-up <agent> to pull prior context before deciding.list, status, team list, task list, memory list/show/search/wake-up/t2-search.send, run, chat, task add/claim/complete, team create/retire, memory put/rm/mine) act on the live fleet — run them only when the user's intent is clear.--json — machine-readable output (available on most verbs) for parsing before you act.--workspace <PATH> — point at a specific workspace root if not auto-resolved.--lang <en|th> — output language.