Back to skill

Security audit

CNC Quote Workflow

Security checks for vulnerabilities and agentic risk

Overview

This CNC quoting skill is mostly coherent, but it deserves review because quote details can be logged or retained and the workflow declares email/calendar notifications without clear user controls.

Review this before installing in a production quoting environment. Keep UniSkill disabled unless you control the referenced local skill path, disable or explicitly configure any email/calendar notifications before handling customer data, and avoid running it where application logs are broadly retained or visible. Pin PyYAML for reproducible installs and remove the automatic report footer if reports will be sent to customers.

Vulnerability Patterns
  • Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
  • Insecure DependenciesIntroduces malicious components through unsafe dependency sources
  • 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 (2)

T01 · Skill Instruction Hijacking

Error
Location
workflow_engine.py:313
Finding

Deterministic Promotional Content Injection into Generated Reports

Content
View full analysis
🦫 Beaver | Reliable, capable, and ready > Author: Timo (miscdd@163.com) ``` ### Technical Analysis `WorkflowEngine.generate_report()` unconditionally appends author branding, a promotional slogan, and an email address to every generated CNC quote report. The documented API provides no option to disable or replace this footer. Because the content is inserted into the trusted final response rather than kept in project metadata, every caller and downstream system using `generate_report()` receives and may redistribute content unrelated to the requested quote. This constitutes deterministic manipulation of the Agent's output channel. ### Attack Path 1. A user or integrating Agent loads the CNC quote Skill. 2. The caller invokes `WorkflowEngine.execute()` to calculate a quote. 3. The caller uses the documented `generate_report()` method to render the result. 4. `generate_report()` unconditionally appends the fixed branding and contact information. 5. The promotional content is returned and may be displayed, stored, or redistributed as part of the trusted quote report. No additional privileges, special input, or user interaction are required. ### Impact Assessment The issue does not provide operating-system access, code execution, or elevated privileges. Its scope is limited to the integrity of generated reports and downstream content derived from them. Potential effects include: - Unrequested promotion in every generated report. - Unintentional redistribution of personal contact information. - Loss of output neutrality and report integrity. - Branding contamination in automated business documents or downstream systems. ]]>
Remediation
View remediation
str: # Generate task-related report content. if include_attribution: report += "\n\nGenerated by CNC Quote Workflow" return report ``` 4. Avoid including personal email addresses in automatically generated business documents. 5. Add a test asserting that default reports contain only user-requested quote information. ]]>

T08 · Insecure Dependencies

Warning
Location
requirements.txt:1
Finding

Unpinned Third-Party Dependency Permits Unreviewed Package Updates

Content
View full analysis
=6.0 ``` Related installation instructions also install the package without a lock file or integrity verification: ```bash pip install pyyaml ``` ### Technical Analysis The dependency declaration accepts any PyYAML version greater than or equal to 6.0. Consequently, separate installations may resolve to different package versions, including future releases that were not included in this audit. `workflow_engine.py` imports `yaml` at process startup, so code supplied by the resolved package executes with the privileges of the Python process. Although the project correctly uses `yaml.safe_load()` and there is no evidence that the current PyYAML package is malicious, the installation process does not provide version reproducibility or package integrity verification. This is a supply-chain hardening weakness rather than evidence of an active dependency-confusion or typosquatting attack. ### Attack Path 1. A user follows the documented installation command or installs `requirements.txt`. 2. The package resolver selects any available PyYAML release satisfying `>=6.0`. 3. The selected package is installed without an audited lock file or required cryptographic hash. 4. `workflow_engine.py` imports the installed `yaml` package at startup. 5. If a future selected release or its distribution channel is compromised, package-controlled code executes under the application user's account. Successful exploitation is contingent on compromise or malicious replacement of an accepted dependency release; no such compromise was identified in the audited repository. ### Impact Assessment A compromised dependency could execute code with the same permissions as the user or service running the workflow. Depending on that environment, this could permit: - Reading files available to the w ...[truncated 331 chars]
Remediation
View remediation
``` 2. Generate and commit a reproducible lock file. 3. Require package hashes during installation, for example by using a hash-pinned requirements file with: ```bash pip install --require-hashes -r requirements.txt ``` 4. Review dependency updates explicitly rather than accepting all future compatible releases. 5. Run dependency vulnerability scanning in CI. 6. Install dependencies from a trusted package index over verified TLS. 7. Execute the workflow under a least-privileged account and restrict unnecessary filesystem and network access to reduce the impact of a dependency compromise. ]]>
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
Findings (19)

Undeclared Tool Scope

Medium
Category
MCP Least Privilege
Confidence
90% confidence
Finding

The skill documents executable components and Python usage but does not declare any explicit tool scope such as permissions or allowed-tools, while static analysis detected file read capability. This creates an authorization ambiguity where the runtime may grant broader filesystem access than users or reviewers expect, increasing the risk of unintended data exposure if any agent or workflow component reads local files.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

The parser logs raw user input verbatim, which can expose sensitive customer data such as part specifications, deadlines, quantities, and potentially proprietary manufacturing requirements into application logs. In a CNC quoting workflow, these logs may be broadly accessible to operators or retained for long periods, increasing confidentiality and privacy risk.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

This code file contains natural-language descriptions and user-facing output entirely in Chinese, including the module docstrings and console messages. Under the language/locale policy rule, forcing a specific language without user opt-in is a policy concern unless the locale restriction is explicitly documented and justified, which is not present here.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

This code file contains user-facing natural-language descriptions that assume Chinese as the operating language ('元认知审核Agent', 'UniSkill V4自我辩论 + 竞品分析'). The policy requires flagging language/locale constraints when a specific language is forced without user opt-in or justification.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The docstring states UniSkill is disabled by default, but when config.json is missing the function falls through to enabling the optional layer. This insecure-default behavior can silently activate the external code-loading path, expanding attack surface and violating operator expectations about deployment safety.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The code prepends an external filesystem path to sys.path, changing Python import resolution for the running process. In a workflow that should only perform CNC quote review, this creates an unnecessary code-loading capability that can execute attacker-controlled modules from that directory if the path contents are modified or replaced.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

This Python file contains natural-language descriptions entirely in Chinese and presents itself as the skill's operating mode without offering any language choice. Under the policy, forcing a specific language or locale without user opt-in is a natural-language policy violation.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
89% confidence
Finding

The logger message emitted during execution is user-visible natural-language text in Chinese and does not provide any user opt-in or fallback language. This can violate language/locale policy when the skill is used in broader contexts.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
90% confidence
Finding

The completion log and final printed banner include Chinese text as the only language option. Because the file offers no documented language choice or justification for a Chinese-only experience, this is a policy-relevant language constraint.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
80% confidence
Finding

The user-facing name, description, agent labels, output components, and metadata are entirely in Chinese, with no indication that other languages are supported or that Chinese is a documented regional requirement. This can constitute a language-policy issue because it effectively forces a specific language without user opt-in or an explicit justification.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

The workflow enables calendar and email notifications as outbound actions, but the manifest shows no user-facing disclosure, consent flow, recipient constraints, or description of what data may be sent externally. In a quoting workflow that processes customer/job details, this can lead to unintended transmission of sensitive business information or surprise side effects triggered without clear user awareness.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
88% confidence
Finding

The file's user-facing text, docstrings, logs, and generated report are written entirely in Chinese, indicating a fixed language choice. There is no evidence of opt-in, locale selection, or documentation justifying a Chinese-only constraint.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

The workflow logs raw user_input directly via logger.info, which can expose sensitive business data, personal data, or proprietary manufacturing specifications to logs that may be retained, forwarded, or viewed by operators. In this skill context, users are likely to submit quote requests containing confidential part details, making verbatim logging a meaningful data exposure risk rather than a harmless debug artifact.

Content

No source excerpt is available for this finding.

Ssd 3

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The code both logs the raw user_input and stores it verbatim in the result object, creating multiple persistence paths for potentially sensitive natural-language content. Because this workflow processes CNC quote requests, inputs may contain confidential dimensions, materials, deadlines, or customer information, so retaining them unminimized increases exposure through logs, reports, history, debugging, and downstream integrations.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
91% confidence
Finding

This markdown file presents the skill description, usage, and key information only in Chinese, which can amount to a language-policy issue when no user opt-in or alternative language is provided. The file does not state that the skill is intended only for a Chinese-speaking or region-specific audience.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
84% confidence
Finding

The module description and all user-facing strings are exclusively in Chinese, indicating a fixed language/locale behavior. There is no indication that the user can choose another language or that the locale restriction is explicitly documented as a justified regional constraint.

Content

No source excerpt is available for this finding.

Unpinned Dependencies

Low
Category
Supply Chain
Confidence
92% confidence
Finding

The dependency is only minimally version-constrained (pyyaml>=6.0) rather than pinned to an exact known-good version. This can lead to non-reproducible builds and accidental installation of later releases with regressions or newly introduced security issues, especially in automated agent workflows where environments may be rebuilt frequently.

Content

Scanner excerpt · requirements.txt (reported line 1)May include surrounding context.

text
pyyaml>=6.0

# 可选依赖(用于扩展功能)
# ollama>=0.1.0  # 向量嵌入服务

Unverifiable Dependency: pyyaml has 8 known advisory(ies) (CVE-2019-20477 (Deserialization of Untrusted Data in PyYAML); CVE-2020-1747 (Improper Input Validation in PyYAML); CVE-2020-14343 (Improper Input Validation in PyYAML) +5 more), but the manifest does not pin a version, so it is unknown whether the installed release is affected

Low
Category
Supply Chain
Confidence
95% confidence
Finding

PyYAML has a history of security advisories, including unsafe deserialization issues, and the manifest does not pin a version, making it impossible to verify whether deployed environments will use a patched release. In a workflow/agent context, YAML is commonly used for configuration and orchestration, which increases the likelihood that untrusted or semi-trusted YAML may be processed somewhere in the system.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
87% confidence
Finding

The file contains natural-language comments only in Chinese, such as '可选依赖' and other Chinese descriptions, with no indication that language choice is optional or justified by a region-specific purpose. This can violate the language/locale policy because it imposes a specific language on readers without opt-in.

Content

No source excerpt is available for this finding.