Back to skill

Security audit

math-guide-solver

Security checks for vulnerabilities and agentic risk

Overview

The skill is mostly a math-helper package, but it also ships and documents release automation that can modify and publish to GitHub, which is not aligned with ordinary math solving.

Review this before installing. Do not run GITHUB_UPDATE_SCRIPT.sh unless you are the maintainer and have verified the repository, source files, staged diff, and tag policy. Prefer a pinned, verified install source, and avoid uploading sensitive images or documents unless you are comfortable with OCR, LLM, and optional API processing of that content.

Vulnerability Patterns
  • Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
  • Insecure DependenciesIntroduces malicious components through unsafe dependency sources
  • Insecure Skill Coding PracticesFinds exploitable flaws such as hardcoded secrets or command injection
  • 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
Findings (3)

T08 · Insecure Dependencies

Warning
Location
README.md:42
Finding

Mutable Package Execution Through npx

Content
View full analysis
Remediation
View remediation

T03 · Remote Payload Retrieval and Execution

Error
Location
references/QUICKSTART.md:27
Finding

Remote Skill Installation From Mutable and Conflicting Repository Sources

Content
View full analysis
Remediation
View remediation
``` 4. Publish checksums for release archives and require verification before registration. 5. Sign release tags and document the trusted signing key and verification procedure. 6. Pin every Python dependency to an exact reviewed version and provide a hash-locked requirements file. 7. Avoid direct registration from a mutable repository URL. Download, verify, inspect, and only then register the local artifact. 8. Configure repository protection, mandatory review, and publisher-account multi-factor authentication. ]]>

T09 · Insecure Skill Coding Practices

Error
Location
GITHUB_UPDATE_SCRIPT.sh:42
Finding

Release Script Publishes Unverified Shared-Path Content and Can Rewrite Remote Tags

Content
View full analysis
/dev/null 2>&1; then echo "⚠️ Tag v1.1.0 已存在" echo " 是否删除并重新创建?" ...[truncated 3045 chars]
Remediation
View remediation
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Rogue AgentSelf-Modification, Session Persistence
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • Behavioral ASTexec() Call, eval() Call, Dynamic Import
Findings (29)

Description-Behavior Mismatch

High
Category
Not specified by scanner
Confidence
99% confidence
Finding

This script performs repository update, tagging, and remote GitHub push operations even though the skill is described as a math-solving/OCR/rendering tool. That capability mismatch is dangerous because users or automated systems may grant the skill trust appropriate for document/math processing while it can modify source control state and publish releases to a remote repository. In skill ecosystems, hidden or unjustified operational behavior is a strong indicator of supply-chain risk.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

High
Category
Not specified by scanner
Confidence
99% confidence
Finding

The script pushes commits and tags to the remote 'origin' on 'main', which is a powerful source-control modification capability unrelated to solving math problems. If executed in an environment with configured Git credentials, it can alter upstream repository history and publish release artifacts, creating a supply-chain compromise path or unauthorized repository changes.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

The declared description presents an end-user math-processing and problem-solving workflow involving OCR, LaTeX extraction, image rendering, and solution generation. The supplied code does none of that. Instead, it is a release automation script for a GitHub repository: it accesses fixed local filesystem paths, copies documentation files, performs git add/commit/tag operations, optionally deletes an existing tag, and pushes updates to the remote repository. This is a materially different primary purpose with undeclared capabilities involving source control and remote publishing. Therefore the description does not accurately represent the code chunk.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
95% confidence
Finding

The declared description promises a complete workflow centered on image upload, OCR, formula extraction, PNG generation, and detailed solving. The provided code implements only a subset: mode selection for socratic/detailed/quick responses, static hint generation, templated explanation text, and construction of a rendering configuration payload. There is no OCR library use, no image or markdown parsing, no batch processing, no file I/O beyond CLI args, and no actual invocation of a renderer to create PNGs. While the code is aligned with the guided-solution and theming portions of the description, the primary advertised capabilities around extraction and rendering are absent, making the description materially broader than the actual behavior.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
97% confidence
Finding

The description promises a broad end-to-end math workflow centered on image OCR, rendering, and solution guidance. The supplied code does not implement those core capabilities. It only processes a provided text string, performs regex-based formula extraction and notation-to-LaTeX conversion, assigns simple metadata (domain/difficulty), and emits JSON. There is no OCR engine, no image/file ingestion, no rendering to PNG, no explanation or hint generation, and no batch or markdown handling. While LaTeX extraction is partially represented, the primary declared purpose is materially broader and different from the actual code behavior.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

This shell script presents all user-facing instructions, warnings, and prompts in Chinese, including the confirmation flow for tag deletion. That enforces a specific language on users without opt-in, which matches the language/locale policy violation criteria.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The guide instructs the user to execute a shell script (GITHUB_UPDATE_SCRIPT.sh) even though the declared purpose of the skill is math solving, OCR, and rendering. Asking users to run an opaque script from copied/downloaded files creates an unnecessary code-execution path that could perform arbitrary local actions, and the mismatch between capability and purpose makes the instruction especially suspicious in this context.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The guide tells the user to run a shell script without describing its contents, effects, prerequisites, or risks. Encouraging execution of an unreviewed script is dangerous because it can lead to arbitrary code execution, file modification, credential exposure, persistence, or other system compromise, especially when the script is unrelated to the expected math-solver functionality.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

This markdown file presents the skill structure and descriptions entirely in Chinese, including user-facing documentation labels and usage explanations, with no indication that users may choose another language or that the skill is region-specific. That creates a natural-language locale policy concern under the rule for forced language without user opt-in.

Content

No source excerpt is available for this finding.

Rp1

Medium
Category
MCP Rug Pull
Confidence
70% confidence
Finding

npx commands without a version suffix (e.g. @1.0.0) create a rug-pull risk if the upstream server is compromised and publishes a malicious update.

Content

No source excerpt is available for this finding.

Rp1

Medium
Category
MCP Rug Pull
Confidence
89% confidence
Finding

The install command uses npx clawhub@latest, which fetches and executes the latest published package version at runtime rather than a reviewed, pinned release. If the upstream package is compromised or a breaking change is introduced, users could execute untrusted code during installation.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The example trigger phrase "Solve this math problem" is broad and resembles ordinary user language, increasing the chance the skill is invoked unintentionally in contexts where the user did not explicitly mean to call this specific skill. Because this skill processes uploaded content and may forward it to external dependencies, accidental activation can expose user data or cause unintended actions.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
89% confidence
Finding

The override example relies on free-form phrasing ("Solve this with dark theme and detailed explanation") without clearly constraining how the skill should be selected or when option parsing applies. This can cause overbroad matching and accidental activation, especially in conversational environments where similar phrasing is common.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The README lists external dependencies including PaddleOCR and Claude API but does not clearly warn users that uploaded images, extracted formulas, and problem content may be transmitted to third-party services. This is a real privacy and compliance risk because users may upload homework, notes, or images containing sensitive personal or institutional information without informed consent.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The support prompts use vague, everyday phrases like "How do I use this?" and "Show me examples," which are highly likely to appear in unrelated conversations. In an agent ecosystem, such generic prompts can spur unintended routing into this skill, leading to confusing behavior or unnecessary exposure of uploaded math content to external services.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

The skill states integrations with OCR and external APIs but does not clearly warn users that uploaded images and problem content may be transmitted to third-party or internal services for processing. This creates a privacy and data-handling risk, especially if users upload homework, personally identifying information, or sensitive documents embedded in images.

Content

No source excerpt is available for this finding.

Rp1

Medium
Category
MCP Rug Pull
Confidence
93% confidence
Finding

The quickstart instructs users to run npx clawhub@latest install math-solver, which pulls and executes the latest remote package without pinning a specific version. If the upstream package is compromised or a malicious release is published, users could execute attacker-controlled code during installation.

Content

No source excerpt is available for this finding.

Session Persistence

Medium
Category
Rogue Agent
Confidence
60% confidence
Finding

Skill establishes unauthorized persistence across sessions via cron jobs, startup scripts, or state files. Session persistence allows an attacker to maintain access beyond the current interaction.

Content

Scanner excerpt · references/QUICKSTART.md (reported line 56)May include surrounding context.

Set User Preferences

Create ~/.openclaw/skills/math-solver/config.json:

json
{

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The phrase 'Track student progress' indicates collection or retention of learner-related data, which is privacy-relevant behavior. In a markdown skill guide, this should be accompanied by a user warning or disclosure about what is tracked, where it is stored, and whether the user or administrator has control over it.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

The module docstring states it 'Converts images containing math problems,' and the class docstring at L162 describes a processor for math problem images. However, the code never loads image files, performs OCR, handles batches, parses markdown, or renders PNGs; it only takes a text string from argv and extracts regex matches.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The docstring 'Main processor for math problem images' contradicts the behavior of the class, whose public methods accept text and problem_text strings and never operate on image data. This is an active intent mismatch in code documentation rather than merely an omitted detail.

Content

No source excerpt is available for this finding.

Dynamic import via __import__()

Medium
Category
Dangerous Code Execution
Confidence
75% confidence
Finding

Dynamic import() can load arbitrary modules at runtime, bypassing static analysis and potentially importing malicious code.

Content

Scanner excerpt · scripts/process_math_problem.py (reported line 215)May include surrounding context.

python
formulas=formulas,
            domain=domain,
            difficulty=difficulty,
            extracted_at=__import__('datetime').datetime.now().isoformat()
        )
    
    def to_dict(self, problem: MathProblem) -> Dict:

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The trigger phrases are broad and generic, such as 'solve this', 'math problem', and 'homework help', which are likely to match many ordinary user requests beyond the intended narrow invocation scope. This can cause the skill to activate unexpectedly, increasing the chance of inappropriate routing, unnecessary access to uploaded content, or unintended use of dependent OCR/LLM capabilities.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
89% confidence
Finding

Natural-language policy review applies to all file types. The document forces a specific language/locale for all operational instructions, and there is no opt-in, alternative language, or justification that this guide is intentionally region-specific.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

Low
Category
Not specified by scanner
Confidence
83% confidence
Finding

The guide asks the user to discover and disclose the full local path of the math-solver directory and offers to generate files directly into that path. This exceeds what is needed for normal math-solving features and unnecessarily solicits local filesystem information, which could expose sensitive environment details or enable further targeted file manipulation.

Content

No source excerpt is available for this finding.