Back to skill

Security audit

Ai Humanizer.Bak

Security checks for vulnerabilities and agentic risk

Overview

This is a disclosed offline writing-analysis and rewrite skill, but it should not be used to misrepresent who wrote something.

Install only if you want an offline text style analyzer and rewrite assistant. Use it for editing clarity and removing boilerplate, not to hide AI assistance where disclosure, academic integrity, employment, platform, or legal rules require transparency. Avoid making its always-on prompt guidance permanent in shared agents unless users are told that style rewriting is always active.

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
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
  • System Prompt LeakageDirect Leakage, Indirect Extraction, Tool-Based Exfiltration
  • 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
Findings (23)

Tp4

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

There is a clear mismatch between the declared purpose and the actual code. The description claims sophisticated natural-language processing functionality for detecting and rewriting AI-generated text. However, the code chunk only exports an ESLint configuration for JavaScript projects. It sets linting rules and environment options for source and test files and does not process user text, detect AI-writing patterns, compute burstiness/readability metrics, or rewrite prose. The primary purpose of the code is development tooling configuration, which is materially unrelated to the declared skill behavior.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

There is a clear mismatch. The declared description promises a substantial text-processing capability focused on detecting and humanizing AI-generated writing. The actual code only configures the Vitest test framework by enabling globals and defining which test files to include. This is unrelated support/configuration code and does not implement any of the declared functionality.

Content

No source excerpt is available for this finding.

Ae1

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 109)May include surrounding context.

md
echo "Your text here" | node src/cli.js score

Ae1

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 112)May include surrounding context.

md
echo "Your text here" | node src/cli.js score

Ae1

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 115)May include surrounding context.

md
echo "Your text here" | node src/cli.js score

Ae1

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 118)May include surrounding context.

md
echo "Your text here" | node src/cli.js score

Ae1

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 121)May include surrounding context.

md
echo "Your text here" | node src/cli.js score

Ae1

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 124)May include surrounding context.

md
echo "Your text here" | node src/cli.js score

Ae1

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 127)May include surrounding context.

md
echo "Your text here" | node src/cli.js score

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The README explicitly instructs operators to bake a permanent 'humanized' writing style into an agent's system prompt or personality file, making the behavior always-on rather than user-directed. That creates a prompt-level policy override: the skill can silently reshape outputs, suppress normal assistant style/safety phrasing, and reduce user control or transparency about when rewriting is happening.

Content

No source excerpt is available for this finding.

Undeclared Tool Scope

Medium
Category
MCP Least Privilege
Confidence
70% confidence
Finding

Without declared permissions the skill's intent is opaque and cannot be validated.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The invocation language is broad enough that an agent may activate this skill for ordinary editing, polishing, or rewriting requests, not just explicit AI-text humanization. That matters because the skill is designed to alter stylistic markers and make machine-written text appear human-authored, which can facilitate authorship concealment in academic, professional, compliance, or moderation-sensitive contexts.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

The skill explicitly aims to remove signs of AI-generated text and make writing sound as though 'a specific human wrote it,' yet it provides no warning about disclosure, provenance, or integrity risks. In contexts with authorship rules or transparency requirements, this can enable deceptive submission of AI-assisted content and evasion of policy controls.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

The summary logic makes affirmative authorship-style claims such as 'The text looks human-written' and classifies content as 'heavily AI-generated' or 'mostly human-sounding' without clear uncertainty or scope limits. In a skill explicitly designed to detect and rewrite 'AI-like' writing, these outputs can be used to enforce stylistic gatekeeping, misclassify non-native or formulaic human writing, and justify discriminatory moderation or screening decisions.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
90% confidence
Finding

The helper labels encode a normative assumption that high burstiness is 'human-like' and very low burstiness is 'AI-like,' presenting one stylometric feature as a proxy for legitimacy. Because this skill's purpose is to humanize text and score AI detection risk, those labels increase the chance that users or downstream systems will treat dialect, accessibility-oriented, technical, or non-native writing styles as suspicious or inferior.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The style tip explicitly encourages adding first-person phrases such as "I found", "We noticed", and "In my experience" to make text seem more human. In a tool designed to humanize AI text, that guidance can induce users or downstream agents to fabricate personal experience, authorship, or observations that did not occur, which is deceptive and can be harmful in professional, academic, journalistic, or compliance 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 code explicitly labels syllable estimation as an "English heuristic" and strips non-ASCII letters, while other logic also depends on English-centric tokenization and abbreviations. Under the policy rule, this is a locale constraint that affects behavior but is not presented as a user choice or clearly justified as a region-specific tool.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

The comments define certain words and phrases as signals of AI-generated text in a universal way, but the vocabulary includes many normal English register and locale variants, including British spellings such as "endeavour." This encodes a language-style policy preference without offering user choice or documenting that the detector is intended only for a specific English-language context.

Content

No source excerpt is available for this finding.

Unpinned Dependencies

Low
Category
Supply Chain
Confidence
40% confidence
Finding

Dependencies lack version pinning, allowing potential malicious package updates. Consider pinning versions.

Content

Scanner excerpt · package.json (reported line 41)May include surrounding context.

json
"url": "https://github.com/brandonwise/humanizer"
  },
  "devDependencies": {
    "@eslint/js": "^9.39.2",
    "eslint": "^9.0.0",
    "prettier": "^3.0.0",
    "vitest": "^3.0.0"

Unpinned Dependencies

Low
Category
Supply Chain
Confidence
40% confidence
Finding

Dependencies lack version pinning, allowing potential malicious package updates. Consider pinning versions.

Content

Scanner excerpt · package.json (reported line 42)May include surrounding context.

json
},
  "devDependencies": {
    "@eslint/js": "^9.39.2",
    "eslint": "^9.0.0",
    "prettier": "^3.0.0",
    "vitest": "^3.0.0"
  },

Unpinned Dependencies

Low
Category
Supply Chain
Confidence
40% confidence
Finding

Dependencies lack version pinning, allowing potential malicious package updates. Consider pinning versions.

Content

Scanner excerpt · package.json (reported line 43)May include surrounding context.

json
"devDependencies": {
    "@eslint/js": "^9.39.2",
    "eslint": "^9.0.0",
    "prettier": "^3.0.0",
    "vitest": "^3.0.0"
  },
  "engines": {

Unpinned Dependencies

Low
Category
Supply Chain
Confidence
40% confidence
Finding

Dependencies lack version pinning, allowing potential malicious package updates. Consider pinning versions.

Content

Scanner excerpt · package.json (reported line 44)May include surrounding context.

json
"@eslint/js": "^9.39.2",
    "eslint": "^9.0.0",
    "prettier": "^3.0.0",
    "vitest": "^3.0.0"
  },
  "engines": {
    "node": ">=18.0.0"

Unverifiable Dependency: vitest has 3 known advisory(ies) (CVE-2026-47429 (When Vitest UI server is listening, arbitrary file can be read and executed); CVE-2026-84373 (Vitest: Path Traversal / Arbitrary File Read via @vitest/mocker Redirect Mock); CVE-2025-24964 (Vitest allows Remote Code Execution when accessing a malicious website while Vit)), but the manifest does not pin a version, so it is unknown whether the installed release is affected

Low
Category
Supply Chain
Confidence
40% confidence
Finding

Dependency has known vulnerabilities (CVEs). Using packages with unpatched security flaws exposes the environment to known exploits.

Content

No source excerpt is available for this finding.