agentCreate

AdvisoryAudited by Static analysis on Apr 30, 2026.

Overview

No suspicious patterns detected.

Findings (0)

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.

What this means

A wrong target, wrong path, or failed edit could delete an agent workspace, remove channel account settings, or corrupt local OpenClaw configuration.

Why it was flagged

The complete-uninstall path force-deletes the agent/workspace and then directly edits the local OpenClaw configuration at a hard-coded user path to remove channel accounts. This is purpose-related, but it is a high-impact operation with limited containment in the shown snippet.

Skill content
bash "$SKILL_DIR/scripts/openclaw-mac.sh" agents delete {agentId} --force ... path = '/Users/honor/.qclaw/openclaw.json' ... del accounts[account_id]
Recommendation

Use explicit user confirmation immediately before deletion, resolve the config path dynamically, create a backup in the executable workflow, and prefer a scoped supported CLI/API for account removal if available.

What this means

Users may provide bot credentials that allow the created agent to receive or send messages through the configured channel.

Why it was flagged

Optional channel binding may collect and write provider credentials into OpenClaw configuration. This is expected for creating a Feishu bot account, and the artifacts do not show credential exfiltration.

Skill content
"appId": "{appId}",
          "appSecret": "{appSecret}"
Recommendation

Use least-privileged channel credentials, avoid sharing secrets in chat history where possible, and remove or rotate credentials for agents that are no longer needed.

What this means

The actual CLI behavior depends on an external local wrapper that was not reviewed here.

Why it was flagged

The skill relies on another skill's wrapper script, but that helper is not included in this artifact set or pinned by an install specification.

Skill content
所有 CLI 操作必须通过 qclaw-openclaw skill 的 wrapper 脚本执行,禁止直接调用 `openclaw` 命令。
Recommendation

Verify that the referenced qclaw-openclaw wrapper is installed from a trusted source and matches the expected OpenClaw version before using this skill.

NoteHigh Confidence
ASI08: Cascading Failures
What this means

If cleanup is partial, messages intended for a deleted or unbound bot could be handled by the main agent instead.

Why it was flagged

The artifact itself notes that incomplete cleanup can route messages for a removed/unbound channel account to the main agent.

Skill content
若账号配置残留但无 binding,消息会 fallback 到 `defaultAccount` → main agent 响应。
Recommendation

After uninstalling, verify both agent lists and channel bindings, and confirm that unused channel accounts were removed only when not shared by another agent.

NoteHigh Confidence
ASI10: Rogue Agents
What this means

Created agents may continue to exist, keep separate state, and respond through bound channels until explicitly removed.

Why it was flagged

The skill creates persistent independent agents with their own workspaces, sessions, and models. This is the declared purpose, but it is a form of ongoing agent persistence users should understand.

Skill content
创建或卸载独立的 OpenClaw Agent。每个 Agent 拥有独立工作区、独立会话、独立模型,与主 Agent 完全隔离。
Recommendation

Create only agents you intend to maintain, choose clear IDs, review bindings periodically, and uninstall unused agents with verified backups.