Back to skill

Security audit

本地图表生成引擎

Security checks for vulnerabilities and agentic risk

Overview

The skill mostly describes local chart generation, but it also asks for API keys and a callback URL while claiming no network use, so users should review it before installing.

Install only if you are comfortable treating this as a review-needed skill: do not provide API keys or callback URLs unless the publisher clarifies what service is contacted and what data is sent. Use a virtual environment for matplotlib and expect local files to be written under the documented chart memory directory.

Vulnerability Patterns
  • Insecure DependenciesIntroduces malicious components through unsafe dependency sources
  • 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
Findings (1)

T08 · Insecure Dependencies

Warning
Location
SKILL.md:201
Finding

Unpinned Third-Party Package Installation

Content
View full analysis

Vulnerability Details

File Location: SKILL.md, line 201
Vulnerability Type: Unpinned and integrity-unverified dependency installation
Risk Level: Medium

Vulnerable Code Snippet

bash
pip3 install matplotlib

Technical Analysis

The documented recovery procedure instructs users to install matplotlib without specifying an audited version, dependency lockfile, cryptographic hashes, or trusted package index. Consequently, the package and its transitive dependencies are resolved from mutable repository state at installation time.

This creates a supply-chain risk if an upstream release, transitive dependency, configured package index, or mirror is compromised. It can also cause users to install an unexpected future version that has not been reviewed for compatibility or security. Python package installation may execute build backend or installation-related code with the privileges of the user running pip3.

No evidence establishes that the current matplotlib package is malicious. The issue is the unsafe and non-reproducible installation method documented by the skill.

Attack Path

  1. A user runs the advertised skill functionality and encounters a missing matplotlib dependency.
  2. The user follows the recovery instruction and executes pip3 install matplotlib.
  3. pip resolves the latest available package and transitive dependencies using the user's configured index or mirror.
  4. An attacker who has compromised a resolved release, dependency, index, or mirror supplies a malicious distribution.
  5. Package build or installation logic executes under the account running pip3.
  6. The malicious component gains access to resources available to that account and may remain installed in the affected Python environment.

Exploitation therefore requires compromise or malicious control of a package artifact, dependency, or configured package source; the repository itself contains no embedded ma ...[truncated 671 chars]

Remediation
View remediation

Remediation Suggestions

  1. Pin matplotlib and all transitive dependencies to versions that have been reviewed and tested.
  2. Maintain dependencies in a lockfile generated by a tool that supports reproducible resolution.
  3. Require cryptographic hashes for downloaded distributions, such as with pip install --require-hashes.
  4. Explicitly document and enforce an approved HTTPS package index rather than inheriting arbitrary user or system index configuration.
  5. Install dependencies in a dedicated virtual environment with no administrative privileges.
  6. Prefer prebuilt, verified wheels and disable source builds where operationally appropriate.
  7. Add automated dependency vulnerability scanning and a controlled process for updating pinned versions.
  8. Replace the current instruction with a reproducible installation command backed by a reviewed requirements file, for example:
bash
python3 -m venv .venv
. .venv/bin/activate
python3 -m pip install --require-hashes -r requirements.lock
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • 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
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (6)

Intent-Code Divergence

High
Category
Not specified by scanner
Confidence
98% confidence
Finding

The document claims the chart engine is local-only and does not require cloud services, yet later requires an API key and references API-backed execution. This contradiction is dangerous because users may expose credentials or allow networked behavior under a false assumption of offline-only operation, undermining security review and least-privilege decisions.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

High
Category
Not specified by scanner
Confidence
97% confidence
Finding

The Q&A states that some features require an API key, directly conflicting with multiple assertions that the skill never calls third-party services. Such inconsistent documentation can trick operators into provisioning secrets unnecessarily or overlooking unauthorized egress paths because the skill was marketed as local-first and offline.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

High
Category
Not specified by scanner
Confidence
98% confidence
Finding

Deployment guidance instructs users to configure API credentials, establish connections, and call interfaces, which is incompatible with the stated local-only chart-generation model. This creates a serious documentation integrity problem that can lead to secret exposure, unintended external communications, and incorrect security approvals based on false privacy claims.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The input definition says the skill accepts '输入数据或指令' with generic '附加配置选项' but does not define concrete activation phrases, scope boundaries, or exclusions. In a markdown skill description, this makes it unclear when the skill should activate versus when ordinary user requests should not invoke it.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The skill advertises a callback_url input but provides no implementation details, trust boundary, or warning that using it would send data off-host. In a skill repeatedly described as local-only, this mismatch can mislead users or agents into permitting unexpected outbound transmission or assuming privacy guarantees that do not actually hold.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

Documenting a callback_url parameter without any warning about outbound network transmission hides an important security property from users. In the context of a skill sold as local-only, the omission is more dangerous because operators may not realize that using the field could exfiltrate results or metadata to an external endpoint.

Content

No source excerpt is available for this finding.