一键在 Manus 中运行任何 Skill

bwoc-health

星标0

分支0

更新时间2026年6月28日 07:54

Diagnose the health of a BWOC workspace and its agent fleet from Claude Code. Use when the user wants to check fleet/workspace health, find a stuck or unhealthy agent, see which agents are actually running, read an agent's daemon log, inspect trust profiles, or triage before launching work. Read-only diagnostics that wrap the `bwoc` CLI (shell-out, no server).

安装

用 Codex 或 Claude 帮你安装复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。

在 Manus 中运行

来源

bemindlabs

bemindlabs/bwoc-plugin-claude

打开 GitHub 仓库查看创作者相关仓库

下载

在 Manus 中运行

BWOC Health & Diagnostics

This skill teaches when and how to diagnose a BWOC workspace and its fleet from Claude Code. Everything here is read-only (report-only) and a thin wrapper over the bwoc CLI — there is no server or daemon to manage. Discover exact flags with bwoc <verb> --help if unsure.

When to use

Reach for diagnostics when the user asks to:

"Is the fleet healthy?" / "anything broken?" → bwoc fleet health, bwoc doctor.
"Which agents are actually running right now?" → bwoc sessions.
"Why is agent X stuck / what did its daemon say?" → bwoc log <agent>, bwoc status <agent>.
"Can I trust agent X with this?" → bwoc trust <agent>.

Run these before delegating heavy work or when something looks wrong — they are always safe to run.

Fleet & workspace health (read-only)

bwoc fleet health              # all 7 Aparihāniya-dhamma governance signals (report-only)
bwoc fleet health --json       # structured signals for programmatic triage
bwoc doctor                    # diagnose environment + workspace problems
bwoc doctor --json             # structured findings
bwoc info                      # one-card system status (version, phase, workspace, agents)

bwoc doctor can also self-heal safe issues with --auto — that is a mutating verb, so treat it as such (confirm intent before running against a live workspace).

Per-agent diagnostics (read-only)

bwoc status <agent>            # one agent's health + identity snapshot
bwoc status --all              # full detail for every agent
bwoc sessions                  # running agent sessions (markers + process scan)
bwoc log <agent>               # tail the agent's daemon log (stderr)
bwoc log <agent> --lines 100   # more backlog (check --help for the exact flag)
bwoc trust <agent>             # Kalyāṇamitta-7 trust profile (declared + requiredTrust)
bwoc ping <agent>              # liveness check for a served agent (PING → PONG)

Triage flow

Start broad: bwoc fleet health + bwoc doctor to see workspace-level signals.
Narrow: bwoc status --all to spot any agent that is stopped or has a backed-up inbox.
Confirm runtime: bwoc sessions shows who is actually executing right now.
Root-cause: bwoc log <agent> for the daemon's own stderr on a suspect agent.

Safety

Every verb here is read-only except bwoc doctor --auto — confirm before that one.
Treat a non-empty inbox or stopped status as a finding to report, not to silently fix.
Target a specific workspace with --workspace <path> (or BWOC_WORKSPACE) when the cwd is not inside the intended workspace.

同仓库更多 Skills

同仓库

bwoc-council

bemindlabs/bwoc-plugin-claude

Run formal multi-agent decisions and objective tracking on a BWOC workspace from Claude Code — open a council decision (propose), discuss, vote, and resolve by the plugin's voting model + quorum, plus list/track/report OKR key results. Use when a team needs a governed decision recorded (not just a chat) or wants to track objectives. Wraps the `bwoc` CLI; the live paths need a council/okr-kind plugin installed.

2026-06-280

bwoc-knowledge

bemindlabs/bwoc-plugin-claude

Read and write a BWOC workspace's shared knowledge from Claude Code — deep-memory plus the dated document kinds (notes, retrospectives, research, and custom doc kinds). Use when the user wants to recall prior decisions, capture a note or retro after finishing work, file a research doc, or browse what the workspace already knows. Wraps the `bwoc` CLI (shell-out, no server).

2026-06-280

bwoc-lifecycle

bemindlabs/bwoc-plugin-claude

Manage the lifecycle of BWOC agents from Claude Code — incarnate (new), pause/resume (stop/start), supervise the daemon, inspect the base-project binding (debase), and retire. Use when the user wants to add a new agent, pause or reactivate one, keep a daemon alive, or remove an agent from the workspace. Wraps the `bwoc` CLI (shell-out, no server); the mutating verbs follow the workspace's own gates.

2026-06-280

bwoc-messaging

bemindlabs/bwoc-plugin-claude

Read and send messages on a BWOC agent's inbox from Claude Code — inspect an agent's (or the whole fleet's) inbox, count or tail new envelopes, acknowledge/clear a backlog, append a message, and read daemon logs. Use when the user wants to see what work is queued for an agent, triage a backed-up inbox, or hand off a message. Wraps the `bwoc` CLI (shell-out, no server).

2026-06-280

bwoc-quality

bemindlabs/bwoc-plugin-claude

Run a BWOC workspace's verification gates from Claude Code — backend-neutrality audit of agents (bwoc check), workspace structural validation, registry/orphan pruning, and audit-kind plugin findings. Use after editing or adding an agent, before a release, or when the user wants to confirm the workspace is consistent. Read-mostly; wraps the `bwoc` CLI (shell-out, no server).

2026-06-280

bwoc-fleet

bemindlabs/bwoc-plugin-claude

Coordinate the BWOC agent fleet from Claude Code. Use when the user wants to list/inspect BWOC agents, delegate or send work to a fleet member, run a headless task on an agent, manage Saṅgha teams or their shared task list, or read workspace deep-memory. Wraps the `bwoc` CLI (shell-out, no server).

2026-06-080

name	bwoc-health
description	Diagnose the health of a BWOC workspace and its agent fleet from Claude Code. Use when the user wants to check fleet/workspace health, find a stuck or unhealthy agent, see which agents are actually running, read an agent's daemon log, inspect trust profiles, or triage before launching work. Read-only diagnostics that wrap the `bwoc` CLI (shell-out, no server).

BWOC Health & Diagnostics

When to use

Reach for diagnostics when the user asks to:

"Is the fleet healthy?" / "anything broken?" → bwoc fleet health, bwoc doctor.
"Which agents are actually running right now?" → bwoc sessions.
"Why is agent X stuck / what did its daemon say?" → bwoc log <agent>, bwoc status <agent>.
"Can I trust agent X with this?" → bwoc trust <agent>.

Run these before delegating heavy work or when something looks wrong — they are always safe to run.

Fleet & workspace health (read-only)

bwoc fleet health              # all 7 Aparihāniya-dhamma governance signals (report-only)
bwoc fleet health --json       # structured signals for programmatic triage
bwoc doctor                    # diagnose environment + workspace problems
bwoc doctor --json             # structured findings
bwoc info                      # one-card system status (version, phase, workspace, agents)

bwoc doctor can also self-heal safe issues with --auto — that is a mutating verb, so treat it as such (confirm intent before running against a live workspace).

Per-agent diagnostics (read-only)

bwoc status <agent>            # one agent's health + identity snapshot
bwoc status --all              # full detail for every agent
bwoc sessions                  # running agent sessions (markers + process scan)
bwoc log <agent>               # tail the agent's daemon log (stderr)
bwoc log <agent> --lines 100   # more backlog (check --help for the exact flag)
bwoc trust <agent>             # Kalyāṇamitta-7 trust profile (declared + requiredTrust)
bwoc ping <agent>              # liveness check for a served agent (PING → PONG)

Triage flow

Start broad: bwoc fleet health + bwoc doctor to see workspace-level signals.
Narrow: bwoc status --all to spot any agent that is stopped or has a backed-up inbox.
Confirm runtime: bwoc sessions shows who is actually executing right now.
Root-cause: bwoc log <agent> for the daemon's own stderr on a suspect agent.

Safety

Every verb here is read-only except bwoc doctor --auto — confirm before that one.
Treat a non-empty inbox or stopped status as a finding to report, not to silently fix.
Target a specific workspace with --workspace <path> (or BWOC_WORKSPACE) when the cwd is not inside the intended workspace.