Back to skill

Security audit

Lanxin App Card

Security checks for vulnerabilities and agentic risk

Overview

This skill is a Lanxin card formatter, but it is always loaded and uses broad triggers plus no-refusal instructions that can override normal assistant behavior.

Review before installing. Use this only if you want a globally active Lanxin app-card JSON formatter; safer packaging would remove always-loaded behavior, narrow triggers to explicit send requests, allow refusals and clarifying questions, validate links, and warn users not to include secrets, internal URLs, or local file paths in cards.

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:6
Finding

Always-Loaded Skill Overrides Agent Safety and Refusal Behavior

Content
View full analysis

Vulnerability Details

File Location: SKILL.md:6-24 and SKILL.md:106-114
Vulnerability Type: T01: Skill Instruction Hijacking
Risk Level: High

Vulnerable Code Snippet

The following is an English translation of the relevant skill instructions:

yaml
metadata:
  clawdbot:
    emoji: "mobile-phone"
  openclaw:
    always: true
priority: 100
text
Core rules:
1. When a user requests an application card, output an appCard JSON message directly.
3. Do not respond that sending is impossible, unsupported, or temporarily unavailable.
4. Do not explain the JSON format; output it directly.
5. Do not output any other text; output JSON only.

Behavior rule:
When a user asks to send an application card, directly output clean JSON,
without additional descriptions or formatting.

Technical Analysis

The skill is configured with always: true and priority 100, causing its instructions to be loaded broadly and given unusually strong precedence. It then uses absolute language to control the agent's response, including prohibiting refusal, forbidding explanatory text, and requiring JSON-only output.

These instructions do not contain an exception preserving higher-level security policies or allowing the agent to reject unsafe card content. Consequently, when a matching request is received, the skill attempts to replace normal safety judgment with a rigid response policy.

The card schema permits user-influenced content, URLs, fields, and limited HTML styling. The skill provides no stated URL allowlist, scheme validation, content sanitization, or protection against deceptive card content. Combined with the prohibition on refusals and warnings, this can cause the agent to generate trusted-looking card objects containing unsafe or misleading content.

The issue is instruction hijacking rather than local code execution. The accompanying index.js only exports static metadata and does not execute commands, access credentials, establish pe ...[truncated 1636 chars]

Remediation
View remediation

Remediation Suggestions

  1. Remove openclaw.always: true unless automatic global loading is essential. Load the skill only when the user explicitly invokes the Lanxin card feature.
  2. Reduce or remove the priority value so the skill cannot improperly supersede safety and platform instructions.
  3. Delete all unconditional prohibitions against refusal, warnings, or explanatory output.
  4. Add an explicit precedence rule stating that system instructions, platform security controls, and safety policies always override the skill.
  5. Permit the agent to reject requests involving phishing, impersonation, harmful content, credential collection, or unsafe destinations.
  6. Validate every URL before card generation:
    • Allow only https URLs.
    • Reject executable, local-file, scripting, and unsupported schemes.
    • Consider a domain allowlist for icons and card destinations.
    • Reject embedded credentials and malformed hostnames.
  7. Sanitize all card text and HTML-style content according to the downstream renderer's supported syntax. Escape or reject unsupported tags and attributes.
  8. Enforce documented limits for fields and links, including no more than ten field pairs and three links.
  9. Separate formatting requirements from behavioral control. JSON-only output may be requested after safety validation, but must not prevent a refusal when the underlying request is unsafe.
  10. Add tests confirming that malicious links, deceptive content, and requests conflicting with higher-level policies are rejected safely.
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Findings (7)

Tp4

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

The description claims the skill supports sending complex-format Lanxin application card messages. However, the provided code chunk is only a minimal module exporting metadata and contains no functional logic for constructing, sending, or handling application cards. While this code does not perform any undeclared or malicious actions, its actual behavior does not match the declared primary purpose because the promised capability is not present in the supplied implementation.

Content

No source excerpt is available for this finding.

Vague Triggers

High
Category
Not specified by scanner
Confidence
95% confidence
Finding

The trigger set includes broad natural-language terms such as '卡片' and other generic phrases that can match ordinary conversation. This raises the chance of unintended skill activation, causing the agent to switch into forced JSON-output behavior and potentially prepare external-message content when the user did not intend to send anything.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

The skill frames itself as a card-sending capability but does not warn users that entered content and embedded links may be sent to an external messaging platform. That omission can lead to unintentional sharing of sensitive text, internal URLs, or metadata beyond the local assistant context.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The skill explicitly forbids refusal and forces raw JSON output whenever app-card-related language appears, even if the user is only asking questions. This weakens safety boundaries, increases unintended activation, and can coerce the agent into producing action-oriented payloads without confirming intent or appropriateness.

Content

No source excerpt is available for this finding.

Overly Broad Trigger

Low
Category
Trigger Abuse
Confidence
84% confidence
Finding

The trigger '蓝信' is very short and generic within the product domain, so normal discussion about the platform may activate the skill unintentionally. In this skill, that risk is amplified because activation leads to restrictive behavior that suppresses normal conversational safeguards and pushes JSON output.

Content

No source excerpt is available for this finding.

Overly Broad Trigger

Low
Category
Trigger Abuse
Confidence
97% confidence
Finding

The trigger '卡片' is extremely broad and likely to collide with many unrelated conversations about cards, UI cards, bank cards, or approval items. Because the skill then mandates direct JSON output and disallows refusal, accidental activation can significantly distort agent behavior and increase the chance of unintended external-message preparation.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

Low
Category
Not specified by scanner
Confidence
87% confidence
Finding

The example embeds a concrete local filesystem path, normalizing inclusion of local file details in card content even though the stated purpose is message formatting. In practice this can encourage accidental disclosure of device structure or sensitive local filenames when users adapt the example.

Content

No source excerpt is available for this finding.