Back to skill

Security audit

humanizer-es

Security checks for vulnerabilities and agentic risk

Overview

This is a Spanish text-editing skill with no bundled executable code, but its README recommends an unpinned npx install command users should treat carefully.

Before installing, prefer a reviewed tagged release or pinned commit instead of the README's unpinned npx command. Do not run the installer with elevated privileges, and use the skill only on text you are comfortable letting the agent read and edit.

Vulnerability Patterns
  • Insecure DependenciesIntroduces malicious components through unsafe dependency sources
  • 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
Findings (1)

T08 · Insecure Dependencies

Warning
Location
README.md:28
Finding

Recommended Installation Command Executes Unpinned Third-Party Code

Content
View full analysis

Vulnerability Details

File Location: README.md, lines 28-32
Vulnerability Type: T08: Insecure Dependencies
Risk Level: Medium

Vulnerable Code

markdown
### Opción 1: instalación con `npx` (recomendada)

```bash
npx skills add https://github.com/mattc95/Humanizer-es.git
text

### Technical Analysis

The recommended installation method invokes the `skills` npm package through `npx` without specifying an exact package version. If the package is not already available locally, `npx` can retrieve and execute the version currently resolved by the configured npm registry.

The Git repository supplied to the installer is also referenced through its mutable repository URL rather than an immutable commit. Consequently, neither the installer implementation nor the skill content processed by it is cryptographically or immutably tied to the version reviewed during this audit.

This creates a supply-chain trust boundary: future changes to the resolved npm package, compromise of its publisher or registry distribution path, or malicious changes to the repository could cause the documented command to execute behavior that was not present during review.

No evidence was found that the current project itself contains a malicious package or payload. The risk arises from recommending execution of mutable, externally controlled dependencies.

### Attack Path

1. An attacker compromises the npm package account, package distribution process, or another component involved in resolving the unpinned `skills` package.
2. Alternatively, an attacker gains control over the referenced Git repository or introduces malicious content into its default branch.
3. A user follows the installation command recommended in `README.md`.
4. `npx` resolves and executes the externally supplied installer package, which then processes content from the mutable Git repository.
5. Malicious installer or repository behavior runs with the permissi
...[truncated 956 chars]
Remediation
View remediation

Remediation Suggestions

  1. Pin the skills npm package to a specific, reviewed version instead of relying on the registry's current resolution:

    bash
    npx skills@<reviewed-version> add <immutable-source>
    
  2. Reference an immutable Git commit rather than the repository's mutable default branch. Document the expected commit hash and verify that the selected installer supports immutable Git references.

  3. Publish and document checksums or signed release artifacts so users can verify the downloaded skill before installation.

  4. Prefer manual installation as the primary option when feasible. Instruct users to download a tagged release, verify its checksum or signature, inspect the included files, and then copy only SKILL.md and README.md into the destination.

  5. Document the npm package publisher, expected package version, repository commit, files written during installation, and required permissions.

  6. Recommend running the installer in a restricted environment without unnecessary credentials or sensitive environment variables. Users should not run the command with sudo or an administrator account.

  7. Re-audit the pinned installer package and repository revision whenever either dependency is updated.

Vulnerability Patterns
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (4)

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.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
89% confidence
Finding

La frase de activación "editar o revisar texto para eliminar rastros de escritura de IA" mezcla una intención específica con verbos muy amplios como "editar" y "revisar", que aparecen en peticiones cotidianas. No incluye ejemplos negativos ni límites claros sobre cuándo debe activarse frente a una revisión general de estilo.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

El nombre y las instrucciones están completamente orientados a español ("humanizer-es" y todo el contenido operativo en español), pero no se indica que el usuario pueda elegir otro idioma ni que la restricción sea una limitación regional justificada. Según la política indicada, forzar un idioma sin opt-in constituye una infracción de idioma o locale.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
84% confidence
Finding

El título y la descripción presentan la skill como orientada exclusivamente a textos en español, y el documento no ofrece elección de idioma ni explica un mecanismo de opt-in más allá de asumir ese idioma por defecto. Según la política indicada, forzar un idioma o locale sin opción explícita puede constituir una violación de política lingüística.

Content

No source excerpt is available for this finding.