Back to skill

Security audit

Fund Proposal Assistant

Security checks for vulnerabilities and agentic risk

Overview

This skill is a Chinese grant-proposal writing helper with a disclosed optional local script that creates checklist reports and dated document copies.

Before installing, users should know this is primarily Chinese-language grant-writing guidance plus an optional personal-style Windows script. Review and change the hardcoded paths and names before running daily_check.py, and do not rely on it to insert comments into Word documents because it only copies the file and writes a separate Markdown checklist.

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
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Findings (10)

Tp4

High
Category
MCP Tool Poisoning
Confidence
95% confidence
Finding

There is a clear mismatch between the declared purpose (grant-writing support) and the documented behavior of daily automation, version management, and local filesystem modification. This is dangerous because users and orchestrators may invoke the skill expecting text assistance, while it can drive persistent local actions and document copying that were not transparently declared.

Content

No source excerpt is available for this finding.

Undeclared Tool Scope

Medium
Category
MCP Least Privilege
Confidence
87% confidence
Finding

The skill advertises document-writing assistance but also references a local automation script that creates directories, writes Markdown reports, and generates versioned document outputs, while declaring no explicit tool scope. This increases the risk that a host agent may grant broader file access than users expect, enabling unintended local file reads/writes under a benign-looking skill.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

The listed invocation examples such as “帮我写立项依据”, “检查基金申请书”, “生成技术路线图”, and “润色创新点” are generic task requests rather than narrowly scoped activation triggers. In normal conversation, users could say these phrases without intending to invoke this specific skill, and the file does not provide exclusion conditions or narrower trigger constraints.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

The top-level docstring frames the script as a daily materials check that generates a new Word document. However, the only Word-document operation is a filesystem copy in L282-L288, while the claimed substantive changes such as adding comments or a modification record appear only in generated report text and are never applied to the .docx file.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
85% confidence
Finding

Natural-language strings and documentation in this file consistently assume Chinese, including the top-level description and all user-facing output. Under the stated policy, forcing a specific language without user opt-in can be a language/locale policy violation unless the restriction is clearly documented and justified.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The report explicitly states that the new daily version adds revision comments and a front-matter modification record while preserving formatting. In reality, create_daily_version only uses shutil.copy2 to duplicate the source file, so those claimed document edits never occur.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
88% confidence
Finding

The manifest describes a writing assistant for research grant applications, implying content-oriented help such as templates, polishing, and formal checks. This script instead creates dated local copies of Word documents and writes Markdown checklist files, which is project file management behavior rather than direct writing-assistance functionality.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
81% confidence
Finding

The natural-language description is explicitly framed as a Chinese-language grant-writing expert, and the rest of the file presents all interaction patterns exclusively in Chinese. There is no statement offering language choice or clarifying that Chinese is required only because the target funding programs are Chinese-language specific.

Content

No source excerpt is available for this finding.

Missing User Warnings

Low
Category
Not specified by scanner
Confidence
82% confidence
Finding

This code creates or overwrites a daily report file in REPORT_DIR using open(..., 'w'), which is a file-write operation. Although the script prints the generated path afterward, there is no confirmation prompt before writing and the docstring/comments do not clearly warn that running the script will create files on disk.

Content

No source excerpt is available for this finding.

Missing User Warnings

Low
Category
Not specified by scanner
Confidence
90% confidence
Finding

The shutil.copy2 call performs a file-write operation that creates a new document in OUTPUT_DIR. The script logs success after the copy, but there is no explicit pre-execution warning or confirmation that running the skill will create dated document versions in a local directory.

Content

No source excerpt is available for this finding.