Security audit

Openclaw Create Agent

Security checks for vulnerabilities and agentic risk

Overview

The skill performs a disclosed OpenClaw/Feishu setup task, but it can expose Feishu app secrets while making persistent routing changes.

Install only if you intend to let the skill change OpenClaw Feishu routing and restart the gateway. Do not pass real Feishu app secrets through --app-secret unless the script is changed to use a protected input path and redact output; rotate any secret that may already have appeared in command history, terminal output, or logs.

A.I.G

Vulnerability Patterns

Insecure Skill Coding PracticesFinds exploitable flaws such as hardcoded secrets or command injection
Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
Agent Memory PoisoningWrites attacker-controlled rules into memory that affect later sessions
Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally

Findings (1)

T09 · Insecure Skill Coding Practices

Warning

Location: scripts/upsert_openclaw_agent.py:32
Finding: Feishu Application Secret Exposure Through Command-Line Arguments and Script Output
Content: View full analysis
\ --routing-mode \ [--account-id ] \ [--app-id ] \ [--app-secret ] \ [--bot-name ] \ [--peer-kind ] \ [--peer-id ] ``` ### Technical Analysis Secrets supplied through command-line arguments may be exposed through shell history, process inspection utilities, command auditing, terminal capture, CI/CD execution records, or supervising automation. Although process-list visibility varies by operating system and security configuration, command-line arguments are not ...[truncated 2207 chars]
Remediation: View remediation

SkillSpector

Vulnerability Patterns

Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Rogue AgentSelf-Modification, Session Persistence
Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration

Findings (6)

Undeclared Tool Scope

Medium

Category: MCP Least Privilege
Confidence: 90% confidence
Finding: The skill performs sensitive file reads and writes to ~/.openclaw/openclaw.json but does not declare any explicit tool scope or permissions boundary. This increases the chance of unintended or over-broad execution because the invoking agent and user are not clearly informed of the filesystem access the skill requires.
Content: No source excerpt is available for this finding.

Session Persistence

Medium

Category: Rogue Agent
Confidence: 89% confidence
Finding: The skill is explicitly designed to create persistent agent state, modify long-lived configuration, update routing bindings, and change session.dmScope. Persistent state changes are security-relevant because a mistaken or malicious invocation can alter future message routing and account behavior beyond the current session.
Content: Scanner excerpt · SKILL.md (reported line 3)May include surrounding context.
md
--- name: openclaw-create-agent description: Create and wire a new OpenClaw agent with a fixed workflow. Use when the user asks to create/add a new OpenClaw agent or says “我要创建一个新的 Agent”, automate multi-agent setup for Feishu, modify ~/.openclaw/openclaw.json channel accounts, bindings, and session.dmScope, or run follow-up commands such as gateway restart and binding verification. --- # OpenClaw Create Agent

Vague Triggers

Medium

Category: Not specified by scanner
Confidence: 86% confidence
Finding: The description contains broad trigger language such as creating/adding agents, Chinese phrasing, config modification, and follow-up operational commands, which can cause the skill to be selected in situations broader than intended. Over-invocation is dangerous here because the skill makes persistent configuration changes and restarts services.
Content: No source excerpt is available for this finding.

Missing User Warnings

Medium

Category: Not specified by scanner
Confidence: 92% confidence
Finding: The skill advertises modifying ~/.openclaw/openclaw.json, updating bindings, changing session.dmScope, and restarting the gateway without an upfront warning that these are system-impacting and persistent actions. A user may trigger the skill expecting setup help but instead cause service disruption or routing changes affecting existing agents.
Content: No source excerpt is available for this finding.

Session Persistence

Medium

Category: Rogue Agent
Confidence: 86% confidence
Finding: This workflow step confirms the skill will create agent runtime state and upsert persistent config before restarting the gateway. In context, persistence is more dangerous because the skill affects routing and service behavior for future interactions, not just the current request.
Content: Scanner excerpt · SKILL.md (reported line 43)May include surrounding context.
md
1. Align scenario (new bot per agent or multi-group per agent). 2. For multi-group scenarios, run guided gates in Step 0.5. 3. Read current config and backup. 4. Create agent runtime with OpenClaw CLI. 5. Upsert config for channel account, binding, and `dmScope`. 6. Restart gateway and verify. 7. Return a concise change summary.

Natural-Language Policy Violations

Medium

Category: Not specified by scanner
Confidence: 95% confidence
Finding: The skill mandates a fixed Chinese prompt ('Ask this first') for scenario alignment, but it does not offer a language choice or state that Chinese is optional. This can violate language/locale policy because it forces a specific language regardless of the user's preferred language.
Content: No source excerpt is available for this finding.

Static analysis

No suspicious patterns detected.