Security audit

OpenClaw Refactoring

Security checks for vulnerabilities and agentic risk

Overview

This refactoring skill is not clearly malicious, but it can rewrite or restore files in ways that are broader and less safe than its documentation says.

Review before installing. Use only on a backed-up Git working tree, run dry-run commands first, avoid broad roots, and do not use undo with any backup ID you did not verify. Treat the advertised extract/inline/move/batch features and safety claims as unreliable until the implementation is fixed.

A.I.G

Vulnerability Patterns

Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs
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

Findings (3)

T09 · Insecure Skill Coding Practices

Error

Location: scripts/rename.py:93
Finding: Incorrect AST Source Offsets Can Corrupt Files During Rename Operations
Content: View full analysis
Remediation: View remediation

T09 · Insecure Skill Coding Practices

Warning

Location: scripts/main.py:46
Finding: Destructive Rename Is the Default Despite Documentation Claiming Default Preview
Content: View full analysis
Remediation: View remediation

T05 · Unauthorized Access and Privilege Escalation

Error

Location: scripts/rename.py:251
Finding: Unvalidated Backup Identifier Allows Restoration from Arbitrary Directories
Content: View full analysis
Dict[str, Any]: """Undo a refactoring operation""" if backup_id is None: # Find most recent backup if not self.backup_dir.exists(): return {'success': False, 'message': 'No backups found'} backups = sorted(self.backup_dir.iterdir(), key=lambda x: x.stat().st_mtime, reverse=True) if not backups: return {'success': False, 'message': 'No backups found'} backup_path = backups[0] else: backup_path = self.backup_dir / backup_id if not backup_path.exists(): return {'success': False, 'message': f'Backup {backup_id} not found'} # Restore files from backup restored = 0 for backup_file in backup_path.rglob('*'): if backup_file.is_file(): rel_path = backup_file.relative_to(backup_path) target_path = self.root / rel_path target_path.parent.mkdir(parents=True, exist_ok=True) shutil.copy2(backup_file, target_path) restored += 1 return { 'success': True, 'message': f'Restored {restored} files from backup {backup_path.name}' } ``` ### Technical Analysis The caller-controlled `backup_id` is joined directly to `self.backup_dir` without format validation or a containment check. Python `pathlib` permits traversal components such as `..`; an absolute path can also replace the intended base path during path composition. The resulting path is accepted whenever it exists. The method recursively enumerates every file under that directory and copies it into the project root. It does not verify that the path represents a backup created by this tool, consult a trusted manifest, reject symbolic links, or restrict restored fi ...[truncated 1778 chars]
Remediation: View remediation

SkillSpector

Vulnerability Patterns

MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access

Findings (6)

Tp4

High

Category: MCP Tool Poisoning
Confidence: 94% confidence
Finding: This finding indicates the skill may only implement a narrow rename workflow while claiming to be a broad, safe refactoring assistant with suggestions, batch support, and multiple transformation types. Such overstatement can mislead operators into relying on automation that lacks the promised analysis and safety checks, causing widespread incorrect edits across a codebase.
Content: No source excerpt is available for this finding.

Tp4

High

Category: MCP Tool Poisoning
Confidence: 94% confidence
Finding: This finding indicates the skill may only implement a narrow rename workflow while claiming to be a broad, safe refactoring assistant with suggestions, batch support, and multiple transformation types. Such overstatement can mislead operators into relying on automation that lacks the promised analysis and safety checks, causing widespread incorrect edits across a codebase.
Content: No source excerpt is available for this finding.

Tp4

High

Category: MCP Tool Poisoning
Confidence: 96% confidence
Finding: This finding indicates the skill may only implement a narrow rename workflow while claiming to be a broad, safe refactoring assistant with suggestions, batch support, and multiple transformation types. Such overstatement can mislead operators into relying on automation that lacks the promised analysis and safety checks, causing widespread incorrect edits across a codebase.
Content: No source excerpt is available for this finding.

Undeclared Tool Scope

Medium

Category: MCP Least Privilege
Confidence: 89% confidence
Finding: The skill advertises and documents operations that read and modify project files, but it declares no explicit tool scope or permissions boundary. In an agent environment, missing scope metadata can allow the skill to be invoked with broader-than-expected file access, increasing the chance of unintended or unsafe repository modifications.
Content: No source excerpt is available for this finding.

Missing User Warnings

Medium

Category: Not specified by scanner
Confidence: 93% confidence
Finding: This code proceeds from preview directly to executing a multi-file rename operation once the command is run without --dry-run. Although it prints the planned changes, there is no interactive confirmation prompt or equivalent user acknowledgment immediately before the irreversible write phase, and the operation modifies user code across the project.
Content: No source excerpt is available for this finding.

Natural-Language Policy Violations

Low

Category: Not specified by scanner
Confidence: 80% confidence
Finding: The skill content switches into Chinese for core usage instructions and feature descriptions, but does not state that the skill is Chinese-only or offer an alternative language/locale choice. This can violate the language/locale policy when users are implicitly forced into a specific language without opt-in.
Content: No source excerpt is available for this finding.

Static analysis

No suspicious patterns detected.