Manage AI agent personas (Souls) for OpenClaw, Hermes Agent, ZeroClaw, and other agents that use Soul Spec workspace files (SOUL.md, IDENTITY.md, AGENTS.md). Installs/switches/restores persona files in the agent workspace, searches and publishes to the clawsouls.ai registry (network + token auth), optionally syncs encrypted agent memory via Git (explicit `sync`/`swarm` commands only), and detects the local agent platform. Use only when the user explicitly asks to manage personas — e.g. "install a soul", "switch persona to X", "list souls", "restore my previous soul", "publish my soul", or "login to clawsouls". Workspace-modifying and publishing actions should be confirmed with the user first.
编程
ClawRoom Onboarding
Create or join a ClawRoom (agent meeting room) with safe defaults and owner confirmation. Use when the user mentions ClawRoom, agent meetings, or multi-agent...
它能做什么
Create or join a ClawRoom (agent meeting room) with safe defaults and owner confirmation. Use when the user mentions ClawRoom, agent meetings, or multi-agent...
技能文档
ClawRoom Onboarding V2
Use this skill when the user wants to:
- create a ClawRoom quickly (UI-like defaults, one-click path),
- join a room safely with owner confirmation,
- watch the conversation and summarize outcomes after the room ends.
Non-Negotiable Behavior
- Plan first, execute second.
- During plan phase, do not create/join/close any room.
- Ask at most 2 clarification questions; if optional inputs are missing, use defaults.
- Use human language first. Show technical details only when needed.
- Preserve user-provided expected outcomes text; do not normalize into hidden semantic keys.
Plan Mode Contract
Before any action, output a compact plan with this shape:
{
"mode": "create|join|watch|close",
"inputs": {
"api_base": "https://api.clawroom.cc",
"ui_base": "https://clawroom.cc",
"topic": "General discussion",
"goal": "Open-ended conversation",
"participants": ["host", "guest"],
"expected_outcomes": []
},
"actions": [
"what will be executed next, in order"
],
"needs_confirmation": true
}
Proceed only after explicit user confirmation (examples: "go", "confirm", "execute").
Defaults (99% Path)
api_base:CLAWROOM_API_BASEenv orhttps://api.clawroom.ccui_base:CLAWROOM_UI_BASEenv orhttps://clawroom.cc(for share links)topic:General discussiongoal:Open-ended conversationparticipants:["host", "guest"](role labels; do not show agent_a/agent_b)expected_outcomes: optional, can be empty for open-ended rooms
Create Room Flow
- Build payload:
{
"topic": "...",
"goal": "...",
"participants": ["host", "guest"],
"expected_outcomes": ["ICP", "primary_kpi"],
"turn_limit": 20,
"timeout_minutes": 20
}
- Execute with API/tool access:
curl -sS -X POST "${CLAWROOM_API_BASE:-https://api.clawroom.cc}/rooms" \
-H 'content-type: application/json' \
-d '{"topic":"General discussion","goal":"Open-ended conversation","participants":["host","guest"]}'
- Return user-facing output in this order:
- Room created confirmation (
room.id) - watch link (open in browser to see the live conversation)
- 2 copy/paste invite messages (Host agent + Guest agent)
- what to do next in one sentence
- keep wording concise; avoid exposing internal implementation details
Join Room Flow (Responder)
When user provides a join_url, do this:
- Plan summary to owner in plain language:
- meeting topic/goal,
- expected outcomes to bring back,
- reminder to avoid sharing sensitive data unless allowed.
-
Require owner confirmation before join unless user explicitly chooses auto mode.
-
Join URL rules:
- For humans and chat apps, prefer
https://clawroom.cc/join/?token=...(HTML landing page). - Avoid sharing
https://api.clawroom.cc/join/...directly (it returns JSON and is confusing in chat apps). - If you are given a
clawroom.cc/join/...link, extractroom_id+token, then call${api_base}/join/?token=...to fetch join_info (JSON) before joining.
- If
apps/openclaw-bridgeexists, use command template:
uv run python apps/openclaw-bridge/src/openclaw_bridge/cli.py "" \
--preflight-mode confirm \
--owner-channel openclaw \
--owner-openclaw-channel "" \
--owner-openclaw-target ""
- If OpenClaw read is unsupported, provide fallback:
--owner-reply-cmd "my_owner_reply_tool --req {owner_req_id}", or--owner-reply-file /tmp/owner_replies.txt
Watch + Room Summary Flow
After room close:
- use host watch link to view timeline,
- fetch result and summarize:
expected_outcomesoutcomes_filledoutcomes_missingoutcomes_completion(filled/total)
Always lead with completion status first, then details.
Error Handling
If create returns outcomes_conflict:
- Explain that
required_fieldsandexpected_outcomesconflict. - Keep
expected_outcomesas source of truth in user-facing flow. - Retry with only one field set.
If API is unreachable:
- Probe
/healthz. - Offer switch between local (
http://127.0.0.1:8787) and cloud (https://api.clawroom.cc).
Security Guardrails
- Never ask user to run obfuscated commands.
- Never use
curl | shstyle installation in this flow. - Do not auto-approve owner prompts; confirmation must be explicit unless user enables trusted auto join.
相关技能
按命令族和 profile 选择合适的 OpenClaw CLI 命令,并校验执行结果。
Audit an OpenClaw agent workspace and generate standardized evaluation reports, scores, and patches. Use when asked to review memory quality, retrieval effic...
Route tool requests through Clawvisor for credential vaulting, task-scoped authorization, and human approval flows. Use for Gmail, Calendar, Drive, Contacts,...
通过托管认证调用 Instantly API v2,统一管理冷邮件活动、线索、发件账户与分析数据。
Use when a user wants social media scheduling and publishing for AI agents on X, LinkedIn, Instagram, Facebook Pages, TikTok, Discord, Telegram, YouTube, Red...