Create Agent
ReviewAudited by ClawScan on May 10, 2026.
Overview
The skill mostly matches its create-agent purpose, but it includes an unrelated guide for extracting ClawHub login/API tokens from the browser.
Review the package before installing. The create-agent functionality is mostly coherent, but do not follow the included browser-token extraction guide unless you intentionally need a ClawHub token and can protect it like a password. Prefer official login/API-key mechanisms, run dry-run first, and back up OpenClaw configuration before allowing the skill to modify agents, bindings, or TEAM.md.
Findings (4)
Artifact-based informational review of SKILL.md, metadata, install specs, static scan signals, and capability signals. ClawScan does not execute the skill or run runtime probes.
A copied browser token can act like a password for the user's ClawHub account; mishandling it could allow account access or publishing actions.
This file instructs users to extract ClawHub authentication tokens from browser developer tools or local storage. That is sensitive account/session access and is not part of the stated create-agent function or declared requirements.
查找以下字段之一： Authorization: Bearer eyJhbGc... ... Cookie: token=eyJhbGc... ... x-api-key: sk-...
Remove this token-harvesting guide from the skill package or replace it with an official, scoped login/API-key flow. Users should avoid copying browser session tokens unless they fully understand the risk.
Installing and running the skill can change which agents exist and how messages are routed through channels such as Telegram or Feishu.
The skill intentionally changes agent registry, routing bindings, team documentation, and channel account configuration. This is central to its purpose and disclosed, but it is still high-impact configuration mutation.
自动更新 `openclaw.json`（agents.list + bindings）... 自动更新 `TEAM.md`（团队架构）... 如需要，自动在 `channels.<channel>.accounts` 中添加账号配置。
Use dry-run or review openclaw.json and TEAM.md changes before applying them, and keep a backup of existing OpenClaw configuration.
Information saved in the new agent workspace may be reused in later sessions and affect future agent responses.
The generated AGENTS.md template tells future agents to read persistent memory/task records. This is purpose-aligned for an agent workspace, but persistent context can influence future behavior and may contain sensitive information.
4. 查看 `memory/` - 回顾之前的任务记录
Keep sensitive information out of the generated memory directory unless intended, and periodically review or clear persistent agent memory.
New agents and their workspaces can remain active/available after creation until the user removes them.
The skill creates persistent agent directories and workspaces. This persistence is the stated function of the skill, but users should recognize it changes the ongoing OpenClaw agent environment.
~/.openclaw/agents/<id>/agent/ ... ~/.openclaw/workspace-<id>/ ... memory/ ... .openclaw/
Create only agents you intend to keep, verify their bindings and channel tokens, and use the documented deletion process when an agent is no longer needed.