← Back to Skills Marketplace
9446
Downloads
8
Stars
0
Active Installs
10
Versions
Install in OpenClaw
/install safe-exec
Description
Safe command execution for OpenClaw Agents with automatic danger pattern detection, risk assessment, user approval workflow, and audit logging. Use when agen...
Usage Guidance
Install only if you are comfortable granting this skill command execution with your user privileges. Keep SAFE_EXEC_AUTO_CONFIRM, OPENCLAW_AGENT_CALL, and SAFE_EXEC_DISABLE out of untrusted automation, avoid passing secrets in prompts or SAFEXEC_CONTEXT, do not treat the audit log as tamper-proof, and run risky workflows inside an OS/container sandbox.
Capability Analysis
Type: OpenClaw Skill
Name: safe-exec
Version: 0.3.4
The 'safe-exec' skill is presented as a security tool to intercept and require approval for dangerous shell commands executed by an AI agent. Its documentation (SKILL.md, READMEs, CLAWDHUB_SECURITY_RESPONSE.md) explicitly states no network calls, monitoring, or credential requirements, and the developer has actively removed previously problematic features. However, the core scripts (`scripts/safe-exec.sh` and `scripts/safe-exec-approve.sh`) use `eval "$COMMAND"` to execute commands. This is a critical shell injection vulnerability, as a sophisticated attacker could potentially bypass the risk assessment regexes or manipulate the command string to achieve arbitrary code execution, even for commands deemed 'low' risk or after user approval. Despite the clear intent to provide a security layer, this significant RCE risk makes the skill suspicious.
Capability Assessment
Purpose & Capability
The purpose is coherent and local: classify shell commands, queue risky ones, and log activity. The concern is that this high-impact guard also directly executes raw shell strings for low-risk, disabled, approved, and context-downgraded paths, so weak classification or bypass state can become arbitrary command execution with the user's privileges.
Instruction Scope
The docs disclose agent mode and environment controls, but the approval framing is stronger than the implementation: non-interactive approval skips the confirmation prompt, context keywords can downgrade commands, and broad natural-language enable/install phrases are loose for a command-execution control.
Install Mechanism
Installation is a disclosed GitHub clone with chmod and optional PATH symlinks, plus a jq dependency. I found no hidden installer, runtime network calls, credential request, or obfuscated setup behavior.
Credentials
The declared local writes fit the stated purpose and no network or credential use was found. However, environment variables can disable protections or trigger auto-confirm behavior, and SAFEXEC_CONTEXT forwards user context into child process environment and may be logged.
Persistence & Privilege
The skill persists pending requests, rules, and audit logs under ~/.openclaw and executes commands with the invoking user's shell privileges. There is no background daemon in the inspected artifacts, but the approval helper can execute queued commands non-interactively and the audit logs are ordinary local files.
How to Use
- Make sure OpenClaw is installed (local or Docker)
- Run the install command in chat:
/install safe-exec - After installation, invoke the skill by name or use
/safe-exec - Provide required inputs per the skill's parameter spec and get structured output
Version History
v0.3.4
fix: 修复YAML metadata格式以符合ClawHub要求
v0.3.3
Security enhancement
v0.3.2
Add backward compatibility symlinks and improved documentation. Project restructuring with zero breaking changes - all existing users can upgrade without any modifications. New conversational installation method added.
v0.3.1
Remove markdown formatting from description, ensure English-only text
v0.3.0
Add quick install guide and GitHub issue link in description
v0.2.8
Clarify platform-agnostic design - works via in-session terminal notifications, independent of communication tools
v0.2.7
Fully English documentation with global enable mode (Enable SafeExec)
v0.2.6
Update usage to global enable mode (开启 SafeExec) instead of per-command execution
v0.2.5
Update documentation to English with conversational usage examples
v0.2.4
Fix non-interactive environment hang issue
Metadata
Frequently Asked Questions
What is Safe Exec?
Safe command execution for OpenClaw Agents with automatic danger pattern detection, risk assessment, user approval workflow, and audit logging. Use when agen... It is an AI Agent Skill for Claude Code / OpenClaw, with 9446 downloads so far.
How do I install Safe Exec?
Run "/install safe-exec" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.
Is Safe Exec free?
Yes, Safe Exec is completely free (open-source). You can download, install and use it at no cost.
Which platforms does Safe Exec support?
Safe Exec is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).
Who created Safe Exec?
It is built and maintained by OTTTTTO (@ottttto); the current version is v0.3.4.
More Skills