Back to skill

Security audit

Clawpilot

Security checks for vulnerabilities and agentic risk

Overview

Clawpilot is a local, recommendation-only skill advisor; its broad Chinese triggers can overmatch, but it does not install, mutate, exfiltrate, or run hidden payloads.

Install this only if you want a Chinese-oriented advisor for choosing other skills. Review recommendations yourself, especially for legal, medical, mental-health, shopping, and travel categories, because the skill may trigger from broad keywords and its risk labels are manually maintained.

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
Vulnerability Patterns
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • 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 (12)

Vague Triggers

High
Category
Not specified by scanner
Confidence
97% confidence
Finding

The trigger list includes many generic, everyday phrases such as emotional states and common tasks, which can cause the skill to activate in contexts far beyond explicit installation-advice requests. Because this skill influences what other skills a user may install, accidental invocation can expose users to unsolicited recommendations and create an unsafe routing point into higher-risk skills.

Content

No source excerpt is available for this finding.

Undeclared Tool Scope

Medium
Category
MCP Least Privilege
Confidence
70% confidence
Finding

Without declared permissions the skill's intent is opaque and cannot be validated.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

The trigger guidance is broad enough to activate on many ordinary requests whenever a user describes a task and asks for help, which can cause the skill to intercept conversations outside narrow skill-discovery intent. In a skill-selection context, over-triggering can misroute users, surface unnecessary recommendations, and expose them to higher-risk skills or advice flows they did not explicitly request.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

The food-delivery intent includes very broad everyday terms like "吃饭" and "饿了", which can match ordinary conversation rather than an explicit request to use a delivery-related skill. In an advisor skill that recommends or influences installation order, this can cause unintended routing or inappropriate skill suggestions, especially if downstream logic treats keyword hits as sufficient evidence of intent.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The travel/transport intent uses highly generic keywords such as "出行" and "旅行", which overlap with many benign discussions and broad planning conversations. Because Clawpilot is designed to recommend skills, ambiguous triggers can steer users toward transport, booking, or ride-hailing skills without clear user intent, increasing the chance of unwanted recommendations in a sensitive installation-advice context.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The learning/growth intent contains generic words like "学习", "成长", and "技能" that commonly appear in normal conversation and can easily overmatch unrelated requests. In this skill, overbroad educational or self-improvement triggers may result in misleading recommendations, including overlap with higher-sensitivity areas like mental health where the map already contains adjacent skills such as reflection and burnout-related tools.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

This manifest-style JSON lists all user-facing display names, descriptions, and categories in Chinese only, with no indication that users can opt into another language or locale. Because SQP-3 applies to all file types and covers language/locale policy violations, this is a natural-language policy concern rather than a code-security issue.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

The module title, description, comments, CLI usage text, and generated user-facing output are all written exclusively in Chinese, indicating the skill is designed to operate in a single language. There is no indication that users may opt into another language or that the Chinese-only behavior is required for a region-specific purpose.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
78% confidence
Finding

The manifest uses Chinese-only trigger phrases and interaction assumptions without documenting language selection, fallback behavior, or user opt-in for locale-specific routing. This can misroute or exclude users and may trigger the skill based on language-specific heuristics the user did not knowingly choose, though it is primarily a safety/usability issue rather than a direct exploit path.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
90% confidence
Finding

This code's natural-language documentation instructs usage only in Chinese, and the embedded test cases are exclusively Chinese-language queries and outputs. That creates a locale-specific constraint without offering a user choice or documenting why the skill is intentionally region-specific.

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 states its intent map is strongest for Chinese-language queries and frames triggering around shipped Chinese prompts without a clear user opt-in, which can bias routing and responses based on language rather than explicit preference. While not directly enabling code execution or data theft, it can degrade user control and lead to inappropriate activation or lower-quality handling for users who did not choose that mode.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
75% confidence
Finding

All intent names, keywords, and user-facing default response examples are in Chinese, which suggests the skill routing is constrained to a single language/locale. There is no visible opt-in, alternative language option, or documented locale justification in this file.

Content

No source excerpt is available for this finding.