T08 · Insecure Dependencies
Warning
- Location
README_EN.md:110- Finding
Unpinned and Unnecessary Third-Party Dependencies Increase Supply-Chain Exposure
- Content
View full analysis
- Remediation
View remediation
Security audit
humanize-mba-text-skill
Security checks for vulnerabilities and agentic risk
Overview
This skill is locally scoped, but it is explicitly designed to help MBA thesis text reduce AI/plagiarism detection signals and includes a risky auto-fix overwrite edge case.
Review carefully before installing. Use it only for transparent editing, formatting, and citation-quality checks on work you are allowed to revise; do not use it to misrepresent AI-generated or copied academic work as original. If installed, avoid the optional unpinned pip install unless you create an isolated environment, and avoid --apply on non-.txt files unless the overwrite bug is fixed or backups are in place.
Vulnerability Patterns
- 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
- Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
Findings (2)
T09 · Insecure Skill Coding Practices
Warning
- Location
scripts/feedback_generator.py:748- Finding
Auto-Fix Operation Can Silently Overwrite Non-TXT Source Documents
- Content
View full analysis
- Remediation
View remediation
Vulnerability Patterns
- Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
- 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
- Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Findings (109)
Ssd 1
High
- Category
- Not specified by scanner
- Confidence
- 99% confidence
- Finding
The skill is explicitly described as detecting and removing AI-writing traces so thesis text appears more human. In academic contexts, this is directly useful for evading AI-authorship detection and undermining integrity controls, making the skill more dangerous than a generic proofreading tool.
- Content
Ssd 1
High
- Category
- Not specified by scanner
- Confidence
- 99% confidence
- Finding
The documented trigger phrases instruct the agent to remove AI writing traces from supplied text, operationalizing evasion as a first-class user workflow. This lowers friction for misuse and makes the harmful behavior easy to invoke during normal interaction.
- Content
Ssd 1
High
- Category
- Not specified by scanner
- Confidence
- 98% confidence
- Finding
The high-risk strategy explicitly recommends deep rewrites and removal of all AI-feature vocabulary to reduce detection risk. That is not incidental editing guidance; it is a procedural playbook for defeating AI-detection mechanisms in academic submissions.
- Content
Tp4
High
- Category
- MCP Tool Poisoning
- Confidence
- 96% confidence
- Finding
The manifest frames the skill as performing rewriting or polishing, while the described workflow mainly detects AI traces, scores risk, and produces a structured modification plan instead of directly applying humanizing edits. The gap matters because users may unknowingly submit sensitive academic text to a profiling pipeline designed to support concealment-oriented editing rather than straightforward assistance.
- Content
Tp4
High
- Category
- MCP Tool Poisoning
- Confidence
- 99% confidence
- Finding
The manifest frames the skill as performing rewriting or polishing, while the described workflow mainly detects AI traces, scores risk, and produces a structured modification plan instead of directly applying humanizing edits. The gap matters because users may unknowingly submit sensitive academic text to a profiling pipeline designed to support concealment-oriented editing rather than straightforward assistance.
- Content
Tp4
High
- Category
- MCP Tool Poisoning
- Confidence
- 99% confidence
- Finding
The manifest frames the skill as performing rewriting or polishing, while the described workflow mainly detects AI traces, scores risk, and produces a structured modification plan instead of directly applying humanizing edits. The gap matters because users may unknowingly submit sensitive academic text to a profiling pipeline designed to support concealment-oriented editing rather than straightforward assistance.
- Content
Tp4
High
- Category
- MCP Tool Poisoning
- Confidence
- 99% confidence
- Finding
The manifest frames the skill as performing rewriting or polishing, while the described workflow mainly detects AI traces, scores risk, and produces a structured modification plan instead of directly applying humanizing edits. The gap matters because users may unknowingly submit sensitive academic text to a profiling pipeline designed to support concealment-oriented editing rather than straightforward assistance.
- Content
Tp4
High
- Category
- MCP Tool Poisoning
- Confidence
- 99% confidence
- Finding
The manifest frames the skill as performing rewriting or polishing, while the described workflow mainly detects AI traces, scores risk, and produces a structured modification plan instead of directly applying humanizing edits. The gap matters because users may unknowingly submit sensitive academic text to a profiling pipeline designed to support concealment-oriented editing rather than straightforward assistance.
- Content
Tp4
High
- Category
- MCP Tool Poisoning
- Confidence
- 99% confidence
- Finding
The manifest frames the skill as performing rewriting or polishing, while the described workflow mainly detects AI traces, scores risk, and produces a structured modification plan instead of directly applying humanizing edits. The gap matters because users may unknowingly submit sensitive academic text to a profiling pipeline designed to support concealment-oriented editing rather than straightforward assistance.
- Content
Tp4
High
- Category
- MCP Tool Poisoning
- Confidence
- 99% confidence
- Finding
The manifest frames the skill as performing rewriting or polishing, while the described workflow mainly detects AI traces, scores risk, and produces a structured modification plan instead of directly applying humanizing edits. The gap matters because users may unknowingly submit sensitive academic text to a profiling pipeline designed to support concealment-oriented editing rather than straightforward assistance.
- Content
Tp4
High
- Category
- MCP Tool Poisoning
- Confidence
- 98% confidence
- Finding
The manifest frames the skill as performing rewriting or polishing, while the described workflow mainly detects AI traces, scores risk, and produces a structured modification plan instead of directly applying humanizing edits. The gap matters because users may unknowingly submit sensitive academic text to a profiling pipeline designed to support concealment-oriented editing rather than straightforward assistance.
- Content
Tp4
High
- Category
- MCP Tool Poisoning
- Confidence
- 98% confidence
- Finding
The manifest frames the skill as performing rewriting or polishing, while the described workflow mainly detects AI traces, scores risk, and produces a structured modification plan instead of directly applying humanizing edits. The gap matters because users may unknowingly submit sensitive academic text to a profiling pipeline designed to support concealment-oriented editing rather than straightforward assistance.
- Content
Tp4
High
- Category
- MCP Tool Poisoning
- Confidence
- 99% confidence
- Finding
The manifest frames the skill as performing rewriting or polishing, while the described workflow mainly detects AI traces, scores risk, and produces a structured modification plan instead of directly applying humanizing edits. The gap matters because users may unknowingly submit sensitive academic text to a profiling pipeline designed to support concealment-oriented editing rather than straightforward assistance.
- Content
Tp4
High
- Category
- MCP Tool Poisoning
- Confidence
- 99% confidence
- Finding
The manifest frames the skill as performing rewriting or polishing, while the described workflow mainly detects AI traces, scores risk, and produces a structured modification plan instead of directly applying humanizing edits. The gap matters because users may unknowingly submit sensitive academic text to a profiling pipeline designed to support concealment-oriented editing rather than straightforward assistance.
- Content
Tp4
High
- Category
- MCP Tool Poisoning
- Confidence
- 98% confidence
- Finding
The manifest frames the skill as performing rewriting or polishing, while the described workflow mainly detects AI traces, scores risk, and produces a structured modification plan instead of directly applying humanizing edits. The gap matters because users may unknowingly submit sensitive academic text to a profiling pipeline designed to support concealment-oriented editing rather than straightforward assistance.
- Content
Tp4
High
- Category
- MCP Tool Poisoning
- Confidence
- 97% confidence
- Finding
The manifest frames the skill as performing rewriting or polishing, while the described workflow mainly detects AI traces, scores risk, and produces a structured modification plan instead of directly applying humanizing edits. The gap matters because users may unknowingly submit sensitive academic text to a profiling pipeline designed to support concealment-oriented editing rather than straightforward assistance.
- Content
Tp4
High
- Category
- MCP Tool Poisoning
- Confidence
- 98% confidence
- Finding
The manifest frames the skill as performing rewriting or polishing, while the described workflow mainly detects AI traces, scores risk, and produces a structured modification plan instead of directly applying humanizing edits. The gap matters because users may unknowingly submit sensitive academic text to a profiling pipeline designed to support concealment-oriented editing rather than straightforward assistance.
- Content
Tp4
High
- Category
- MCP Tool Poisoning
- Confidence
- 98% confidence
- Finding
The manifest frames the skill as performing rewriting or polishing, while the described workflow mainly detects AI traces, scores risk, and produces a structured modification plan instead of directly applying humanizing edits. The gap matters because users may unknowingly submit sensitive academic text to a profiling pipeline designed to support concealment-oriented editing rather than straightforward assistance.
- Content
Ae1
High
- Category
- analysis-evasion
- Confidence
- 100% confidence
- Finding
Referenced artifact was not completely inspected
- Content
md https://github.com/cangtianhuang/humanizer-academic-zh)（其 `SYSTEM_PROMPT.md` 与 `SKILL.md`）。两份 Skill 的关系：
Description-Behavior Mismatch
High
- Category
- Not specified by scanner
- Confidence
- 99% confidence
- Finding
The file is explicitly framed as a strategy to lower duplicate-check and plagiarism-detection rates, with detailed tactics such as synonym replacement targets, structural rewrites, and quote transformation while preserving meaning. In the context of an MBA-writing humanization skill, this meaningfully expands the skill from style editing into academic-integrity evasion, enabling users to disguise copied or minimally transformed source material.
- Content
Missing User Warnings
High
- Category
- Not specified by scanner
- Confidence
- 97% confidence
- Finding
The document provides extensive misuse-enabling guidance but lacks a strong warning that using rewriting to reduce plagiarism risk or duplicate-check scores can constitute academic misconduct. In this context, the absence of safeguards makes the content more dangerous because the surrounding skill already targets disguising machine-generated or derivative text, increasing the likelihood of deceptive educational use.
- Content
Context-Inappropriate Capability
High
- Category
- Not specified by scanner
- Confidence
- 98% confidence
- Finding
The guidance goes beyond making text sound more natural and instead teaches operational methods for reducing plagiarism-detection rates, including indirect-quote conversion, multi-source merging, and checklist-driven similarity reduction. Because the stated skill purpose is to remove AI-writing traces, these instructions are not necessary for benign use and create a concrete misuse path for academic misconduct.
- Content
Description-Behavior Mismatch
High
- Category
- Not specified by scanner
- Confidence
- 98% confidence
- Finding
The file implements a report generator for AI-likeness detection based on sentence-length CV, which materially differs from the declared skill purpose of rewriting text to reduce AI traces. This capability can be used to evaluate and optimize evasion of AI-detection heuristics, creating an undeclared dual-use function that increases misuse risk and violates the principle of least functionality.
- Content
Description-Behavior Mismatch
High
- Category
- Not specified by scanner
- Confidence
- 98% confidence
- Finding
The file implements AI-trace detection and reporting, while the skill manifest says the skill should humanize/rewrite text to remove AI-writing traces. This capability mismatch is dangerous because an agent may invoke the skill expecting transformation of user content, but instead expose, score, and structure the text for analysis, leading to unintended behavior, privacy leakage in generated reports, and bypass of policy/UX expectations.
- Content
Description-Behavior Mismatch
High
- Category
- Not specified by scanner
- Confidence
- 99% confidence
- Finding
The file's implementation is a detector and rewrite-planning engine for reducing AI-detection signals, not a neutral text-humanization utility. It computes AI-likeness features and then generates targeted modification plans, thresholds, and suggestions explicitly aimed at lowering detection rates, which can enable academic integrity evasion and misuse in deceptive content laundering.
- Content