Security audit

Agent Audit

Security checks for vulnerabilities and agentic risk

Overview

This skill is a disclosed local cost-audit helper with no evidence of exfiltration, persistence, or hidden destructive behavior, though it overstates how much auditing it currently performs.

Install only if you are comfortable letting it read OpenClaw configuration and, if later expanded or run inside an agent session, potentially usage history. Treat its recommendations as limited because the current script mainly summarizes configured agents and model pricing rather than performing the full cron/session audit it advertises. Use --output carefully to avoid overwriting an existing file.

A.I.G

Vulnerability Patterns

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

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 (4)

Undeclared Tool Scope

Medium

Category: MCP Least Privilege
Confidence: 90% confidence
Finding: The skill instructs the user to run a script with an --output flag that writes a report to disk, but the manifest declares no explicit tool scope such as permissions or allowed-tools. That mismatch means the skill has file-write capability without transparent, least-privilege disclosure, which can mislead users and weaken policy enforcement around filesystem access. In this context the write behavior is likely intended for legitimate report generation, but undeclared write capability still expands risk if the script is modified or abused to overwrite arbitrary paths.
Content: No source excerpt is available for this finding.

Intent-Code Divergence

Medium

Category: Not specified by scanner
Confidence: 95% confidence
Finding: The top-level docstring says the script will scan the OpenClaw setup for cost optimization opportunities, and the usage flags suggest history-based auditing. However, the implementation later admits cron/history analysis is not performed here and only outputs a framework plus general guidelines, creating a direct documentation-to-code mismatch.
Content: No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium

Category: Not specified by scanner
Confidence: 98% confidence
Finding: The skill description claims a substantive audit over cron jobs, session history, model usage, waste detection, and ROI optimization. In practice, the main audit path only loads the config file, lists agent models, and explicitly states that detailed recommendations require unavailable cron API access, so the implemented behavior is materially narrower than advertised.
Content: No source excerpt is available for this finding.

Missing User Warnings

Low

Category: Not specified by scanner
Confidence: 87% confidence
Finding: The script writes the generated report to any user-supplied path without confirmation, overwrite checks, or warnings about sensitive contents. In an agent context, this can accidentally overwrite existing files or leave configuration-derived information in an unintended location, increasing local data exposure risk.
Content: No source excerpt is available for this finding.

Static analysis

No suspicious patterns detected.