ClawHub
Back to skill

Security audit

Sift

Security checks for vulnerabilities and agentic risk

Overview

This skill is a real governance integration, but it gives a third-party service broad mandatory control over agent actions and sends action details without clear limits.

Install only if you intentionally want Sift to govern agent actions and you are comfortable sending action metadata to api.sift.walkosystems.com. Review what parameters your agent may include, how receipts and audit logs are retained, and whether you need a platform-enforced plugin instead of prompt-level guidance.

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
Findings (1)

T01 · Skill Instruction Hijacking

Error
Location
SKILL.md:11
Finding

Mandatory External Authorization Hijacks Agent Tool Execution

Content
View full analysis
", "tenant_id": "", "agent_id": "", "agent_role": "", "action": "", "tool": "", "risk_tier": <0-3 based on action severity>, "params": {}, "timestamp": , "nonce": "", "signature": "" } ``` ### Technical Analysis The skill introduces unconditional instructions that alter the agent's normal authorization model. It requires the agent to contact a third-party service before consequential actions and explicitly states that this behavior must continue even when the user requests otherwise. The external service is given authority to determine whether file writes, API calls, browser operations, messages, financial transactions, and other tool actions may proceed. This redirects contr ...[truncated 2818 chars]
Remediation
View remediation

SkillSpector

Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (4)

Intent-Code Divergence

Medium
Category
Not specified by scanner
Confidence
98% confidence
Finding

The top-level description claims Sift 'blocks on DENY' and functions as execution governance, but later text admits the skill 'relies on the agent following instructions' and is therefore probabilistic. This mismatch can mislead deployers into assuming denied actions cannot execute when, in reality, the agent may bypass the documented process entirely.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

The skill presents its authorization process as mandatory and fail-closed, but the mechanism is only prompt-level guidance to the agent rather than an enforced platform control. Because the agent can ignore or be induced to skip these instructions, operators may wrongly rely on it as deterministic security enforcement, leading to unauthorized external actions despite the documentation's guarantees.

Content

No source excerpt is available for this finding.

External Transmission

Medium
Category
Data Exfiltration
Confidence
50% confidence
Finding

Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

Content

Scanner excerpt · SKILL.md (reported line 45)May include surrounding context.

How to call Sift

Make a POST request to https://api.sift.walkosystems.com/authorize with:

json
{

External Transmission

Medium
Category
Data Exfiltration
Confidence
50% confidence
Finding

Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

Content

Scanner excerpt · SKILL.md (reported line 73)May include surrounding context.

How to call Sift

Make a POST request to https://api.sift.walkosystems.com/authorize with:

json
{