Back to skill

Security audit

Lanxin Link Card

Security checks for vulnerabilities and agentic risk

Overview

The skill is a narrow Lanxin link-card formatter, but it is always enabled and tells the agent not to refuse or add warnings when producing link-card JSON.

Review before installing in environments where generated linkCard JSON is automatically delivered or rendered. Prefer a version with narrower Lanxin-specific triggers, explicit user confirmation, URL validation, and permission for the agent to refuse or warn about unsafe or misleading links.

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:20
Finding

Always-Enabled Skill Overrides Refusal and Safety Behavior for Link Requests

Content
View full analysis
` or `` tags** ``` ```markdown ## Strictly Prohibited Incorrect Behavior ❌ Prohibited: missing required fields ❌ Prohibited: malformed JSON ❌ Prohibited: explaining “this is JSON” or “the system will parse it” ❌ Prohibited: refusing to send ❌ Prohibited: outputting only textual content instead of JSON ❌ Prohibited: outputting any other text; output only JSON ❌ Prohibited: wrapping JSON in a Markdown code block ❌ Prohibited: outputting `` or `` tags ❌ Prohibited: outputting Markdown-formatted text ❌ Prohibited: adding any text before or after the JSON ## Behavioral Rules When the user asks you to send a link card → **directly output clean JSON** → do not add any description → do not add formatting wrappers → complete. ``` ### Technical Analysis The skill is configured as always enabled and assigns itself a priority of `100`. It also registers broa ...[truncated 3367 chars]
Remediation
View remediation
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 (5)

Vague Triggers

High
Category
Not specified by scanner
Confidence
95% confidence
Finding

The trigger list contains broad, common phrases such as “链接” and “蓝信” that can activate the skill in many unrelated conversations. This increases the chance of unintended invocation, causing the agent to emit structured message JSON when the user did not explicitly request this action, which can disrupt normal operation or enable message spoofing workflows.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
88% confidence
Finding

The skill mandates a rigid output format and forbids any explanatory or refusal text, regardless of user intent or safety context. This can suppress normal safety handling and force the agent to produce actionable JSON even when clarification, validation, or refusal would be more appropriate.

Content

No source excerpt is available for this finding.

Overly Broad Trigger

Low
Category
Trigger Abuse
Confidence
70% confidence
Finding

Overly Broad Trigger: '蓝信' is too short and may match unintended inputs

Content

No source excerpt is available for this finding.

Overly Broad Trigger

Low
Category
Trigger Abuse
Confidence
97% confidence
Finding

The single-word trigger “链接” is extremely generic and likely to appear in ordinary user requests unrelated to sending a 蓝信 link card. In combination with the skill’s strict instruction to output only JSON, accidental activation could cause unexpected structured output or unauthorized message composition behavior.

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 string is written only in Chinese ("蓝信链接卡片发送技能"), which can constitute a language/locale policy violation if the organization expects skills to avoid forcing a specific language without user opt-in. The file does not offer any alternate language, locale selection, or indication that the skill is region-specific.

Content

No source excerpt is available for this finding.