Back to skill

Security audit

Semantic Scholar Search

Security checks for vulnerabilities and agentic risk

Overview

The skill appears to be a legitimate Semantic Scholar search tool, but its install metadata declares a Python dependency through npm, which could fail or pull an unintended package.

Review or fix the dependency metadata before installing: Python requests should be declared in a Python dependency file or otherwise handled by the skill manager, not as an npm dependency. Also expect search terms to be sent to Semantic Scholar, and note that the visible instructions are primarily in Chinese.

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
package.json:12
Finding

Python Dependency Declared Through the Wrong Package Ecosystem

Content
View full analysis

Vulnerability Details

File Location: package.json, lines 12-14
Vulnerability Type: Cross-ecosystem and unpinned dependency declaration
Risk Level: Medium

Vulnerable Code

json
"dependencies": {
  "requests": ">=2.31.0"
}

Technical Analysis

The project executable is written in Python and imports the Python requests module. However, package.json instructs npm-compatible tooling to resolve a package named requests from the npm registry rather than resolving the intended package from the Python Package Index.

Package names are scoped independently across package ecosystems. An npm package named requests is therefore not equivalent to the Python requests library. Processing this manifest through npm can resolve an unrelated component or fail because no npm release satisfies the specified range. The open-ended >=2.31.0 constraint and absence of a dependency lockfile also prevent reproducible dependency resolution.

If a package manager resolves an unintended package, that package may include npm lifecycle scripts that execute during installation with the permissions of the installing user. The audited repository does not itself contain such a lifecycle script; the risk arises from resolving dependencies through the wrong ecosystem.

Attack Path

  1. A user, deployment service, or Skill manager identifies package.json as the dependency manifest.
  2. It runs an npm installation command in the project directory.
  3. npm attempts to resolve requests from the npm registry rather than obtaining Python's requests package from PyPI.
  4. The installation either fails, affecting availability, or retrieves an unrelated npm package if a matching version becomes available.
  5. Any lifecycle hooks supplied by the resolved npm package can run during installation under the privileges of the user or automation service performing the installation.

Impact Assessment

The immediate likely ...[truncated 643 chars]

Remediation
View remediation

Remediation Suggestions

  1. Remove the Python requests dependency from package.json.
  2. Declare Python dependencies in an ecosystem-appropriate manifest such as requirements.txt or pyproject.toml.
  3. Pin requests to a reviewed, exact version rather than using an unrestricted lower-bound range.
  4. Generate and commit a lockfile with integrity hashes using an appropriate Python dependency-management tool.
  5. Configure deployment automation to install dependencies only from the expected package ecosystem and approved registries.
  6. Disable package lifecycle scripts where they are unnecessary and supported by the installation environment.
  7. Add automated checks that reject Python packages declared as npm dependencies and detect dependency-manifest inconsistencies.
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
Findings (4)

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The skill description and user-facing instructions are written in Chinese, including the manifest description and usage guidance, with no indication that another language can be used or selected. This can violate a language/locale policy when a skill implicitly forces one language without user opt-in.

Content

No source excerpt is available for this finding.

External Transmission

Medium
Category
Data Exfiltration
Confidence
60% confidence
Finding

Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

Content

Scanner excerpt · semantic_scholar.py (reported line 6)May include surrounding context.

python
Semantic Scholar Search - Free academic paper search tool

Based on Semantic Scholar API v2
https://api.semanticscholar.org/
"""

import requests

External Transmission

Medium
Category
Data Exfiltration
Confidence
60% confidence
Finding

Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

Content

Scanner excerpt · semantic_scholar.py (reported line 19)May include surrounding context.

python
Semantic Scholar Search - Free academic paper search tool

Based on Semantic Scholar API v2
https://api.semanticscholar.org/
"""

import requests

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

This code presents the primary user-facing usage instructions in Chinese, while the rest of the tool and identifiers are otherwise not clearly scoped to a Chinese-only audience. That creates a language/locale policy issue because users are not given an opt-in choice or explanation that the skill is intentionally region-specific.

Content

No source excerpt is available for this finding.