Back to skill

Security audit

RiskBlind Radar 风险盲区雷达

Security checks for vulnerabilities and agentic risk

Overview

This is a disclosed Chinese-language decision-risk prompting skill with no executable code, network access, credential handling, or hidden destructive behavior.

Install this only if you want your assistant to proactively challenge plans, commitments, budgets, and assumptions in Chinese. Review or narrow the trigger phrases if you do not want frequent risk-oriented follow-up questions, and enable any optional observation log only if you are comfortable storing those notes locally.

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
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (14)

Vague Triggers

High
Category
Not specified by scanner
Confidence
96% confidence
Finding

The trigger design explicitly states 'triggered then ask' and scans every message for broad topic mentions, which makes the skill prone to over-activation on ordinary conversation. In an agent skill, this can cause persistent unsolicited probing, derail user intent, and create safety and UX issues by steering conversations based on loose keyword matches rather than clear contextual relevance.

Content

No source excerpt is available for this finding.

Vague Triggers

High
Category
Not specified by scanner
Confidence
98% confidence
Finding

The listed trigger words for T1 include very common everyday expressions like 'plan', 'will do', and 'later', which are likely to appear in harmless chat. This can make the skill ask risk-oriented follow-ups too often, effectively hijacking normal dialogue flow and causing the agent to infer decision-making context where none exists.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The skill declares that it auto-triggers on 'specific topics' but does not define those topics in the skill file, making activation boundaries unclear and hard to audit. In a conversational agent, ambiguous triggers can cause the skill to fire in unintended contexts, leading to intrusive questioning, privacy overcollection, or interference with other safety-critical instructions.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

The instruction to 'scan trigger words' is overly broad and provides no constrained vocabulary, matching logic, or collision-avoidance rules. This can cause accidental activation on ordinary language, making the agent behavior unpredictable and potentially steering conversations in ways the user did not request.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
86% confidence
Finding

All user-facing instructions, triggers, and question templates are presented exclusively in Chinese, and the file does not indicate that language use is optional or region-specific. This can violate language/locale policy when a skill implicitly forces one language without user opt-in.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

This section continues the pattern of using broad conversational cues such as hesitation or future-oriented language that can collide with normal speech. While not directly enabling code execution or data exfiltration, it increases the chance of manipulative or noisy agent behavior through repeated, context-insensitive interrogation.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
90% confidence
Finding

The user-assumption triggers rely on phrases like 'users will' or 'users need', which are common in brainstorming and product discussion and may fire without enough context. In this skill's proactive-questioning context, that can push the agent into assertive challenge mode too readily, creating undue pressure or misleading scrutiny.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

The commitment-tracking trigger list includes highly common temporal phrases like 'this week', 'tomorrow', or 'within two days', which appear in many harmless messages. Because this template asks for accountability and consequences, false activations can feel coercive or intrusive, especially if the agent repeatedly treats casual statements as binding commitments.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
85% confidence
Finding

All user-facing instructions and descriptions in the file are presented in Chinese, and there is no indication that the skill supports other languages or requires Chinese for a specific regional or compliance reason. This can violate language-choice policy when a skill implicitly forces one language without user opt-in.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The trigger phrases are intentionally broad and map to very common conversational patterns such as planning, budgeting, waiting, and commitments. In a conversational agent, this can cause the skill to activate unexpectedly during ordinary dialogue, leading to prompt interference, user frustration, and potentially steering decisions when the user did not explicitly request risk analysis. The skill context makes this more plausible because it is designed to proactively interrupt and ask follow-up questions.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
89% confidence
Finding

The opportunity-detection rule activates when a user 'complains but has not turned it into action,' with examples like '每次都...' and '太麻烦了', which are broad enough to match routine frustration. This ambiguity can make the skill aggressively reframe casual complaints into product or monetization opportunities, potentially derailing user intent and prompting unwanted business-oriented probing.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

The '快速触发' table maps very common phrases such as '投钱', '上线', '计划做X', and '用户会...' directly to automatic follow-up questions without clear scope limits or gating criteria. In an agent skill, this can cause over-triggering across many ordinary conversations, leading to intrusive steering, reduced user control, and unintended collection or inference of sensitive business/planning context.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
95% confidence
Finding

All natural-language instructions, triggers, and guidance in this file are presented only in Chinese. If this skill is used in a broader multilingual environment, this effectively imposes a language constraint without explicit user opt-in or justification.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
90% confidence
Finding

The description is written entirely in Chinese, which can indicate a language-specific skill experience without any stated user opt-in or locale selection. Under the policy, language constraints should either offer a choice or be clearly documented as intentional and justified.

Content

No source excerpt is available for this finding.