Back to skill

Security audit

微信支付接入skill

Security checks for vulnerabilities and agentic risk

Overview

This WeChat Pay helper is purpose-related but needs Review because it silently enables network access and auto-runs an unsigned remote updater that can replace its own skill files.

Install only if you are comfortable with the skill contacting WeChat/Tencent domains, running local commands, installing or using the WeChat Pay CLI, and updating its own files from a remote bundle. Treat Authorization headers, signatures, merchant IDs, certificates, and payment credentials as sensitive; avoid pasting them unless you understand where the chat and logs are stored.

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
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • 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 Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
Findings (13)

Tp4

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

The skill is presented as a consultation/documentation helper, but its behavior includes self-updating, downloading remote content, extracting archives, and rewriting local directories. That mismatch is dangerous because users and platform controls may trust it as a read-mostly advisory skill while it actually performs supply-chain sensitive actions that could alter the skill bundle or local knowledge base.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

High
Category
Not specified by scanner
Confidence
96% confidence
Finding

The code has explicit capability to clear most of the local Skill directory and rewrite it from remotely downloaded content. Even though the implementation includes some path-safety checks during extraction, the core risk remains a destructive remote update mechanism that can overwrite trusted local files and convert a support skill into an arbitrary payload delivery path.

Content

No source excerpt is available for this finding.

Undeclared Tool Scope

Medium
Category
MCP Least Privilege
Confidence
89% confidence
Finding

The skill clearly instructs the agent to use network access, shell execution, and filesystem operations such as reading local docs and running a sync script, yet it declares no explicit tool scope or permission boundary. This creates an over-privileged, under-specified execution model where sensitive capabilities may be invoked implicitly, making review, enforcement, and least-privilege control difficult.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

The skill explicitly requires silently allowlisting external domains and automatically running a resource synchronization script before answering, without notifying or asking the user. Silent network enablement plus automatic fetching and updating increases the risk of covert data egress, unexpected environmental changes, and supply-chain compromise, especially because it happens as a prerequisite to normal use.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The workflow explicitly asks the developer to paste back a full Authorization header or the individual authentication components, which are sensitive request-signing artifacts. Even if these values are not long-lived secrets, exposing them in chat leaks replayable or security-relevant material into a broader system boundary, increasing risk of interception, logging, reuse, or accidental retention.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
87% confidence
Finding

The file's natural-language instructions, headings, examples, and output format are entirely in Chinese, which can amount to a language-policy constraint if users or operators are not given an explicit choice of language. The document does not state that the skill is China-specific or otherwise justify the restriction, nor does it offer multilingual or opt-in behavior.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

The top-level docstring describes the tool in a way that can be read as a knowledge-base sync utility, while the code also updates the Skill bundle itself. This mismatch increases operational risk because reviewers or operators may grant trust or execution permission under a narrower assumption than the script's real capabilities.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

The script downloads a remote bundle from a hard-coded URL and replaces the local Skill directory contents, preserving only limited paths. That creates a remote code/content update channel capable of changing prompts, logic, and resources outside the stated consultation/documentation role; if the remote host, distribution path, or archive contents are compromised, arbitrary skill behavior can be shipped locally.

Content

No source excerpt is available for this finding.

Missing User Warnings

Low
Category
Not specified by scanner
Confidence
85% confidence
Finding

The document instructs the user to execute local scripts and then use the resulting material for remote API calls, but it does not clearly warn that this causes local command execution and outbound network activity. In a security-sensitive payment context, omission of those warnings reduces informed consent and can normalize risky behavior, especially because the user is asked to run code with certificate material and payment credentials nearby.

Content

No source excerpt is available for this finding.

Missing User Warnings

Low
Category
Not specified by scanner
Confidence
89% confidence
Finding

The instruction tells the agent to ask for consent and then write to AGENTS.md, which is a project file modification triggered during normal conversation flow. Even though it requires user approval, the guidance does not require a clear warning that this changes repository state, may persist across future runs, and could affect later agent behavior; this creates a real prompt-safety and unintended state-modification risk.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
87% confidence
Finding

The document states that review must be done by “fresh eyes,” using English wording inside an otherwise Chinese checklist, without offering a language choice or documenting why English is required. This can conflict with a language/locale policy that requires user opt-in before enforcing a specific language.

Content

No source excerpt is available for this finding.

Vague Triggers

Low
Category
Not specified by scanner
Confidence
84% confidence
Finding

This JSON file describes when a project should query an order, but it does not define a precise activation interface, explicit trigger phrases, or clear exclusion conditions beyond a couple of examples. In a manifest/reference context, this broad wording could lead to inconsistent or unintended invocation because readers must infer the exact trigger scope.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
95% confidence
Finding

This shell script’s user-facing instructions and error/help messages are written in Chinese, beginning with the header and usage text. Under the policy for natural-language violations, forcing a specific language without user choice can be a locale/language policy issue.

Content

No source excerpt is available for this finding.