Back to skill

Security audit

pubmed-paper-monitor

Security checks for vulnerabilities and agentic risk

Overview

This PubMed skill mostly matches its stated purpose, but it automatically saves reports to the Desktop and sends a hard-coded personal email address to NCBI, so it should be reviewed before installation.

Install only if you are comfortable with PubMed queries being sent through NCBI using the bundled script's hard-coded email address and with large result sets being saved automatically on the Desktop. Prefer a version that lets you configure the Entrez email, choose whether and where to save reports, and pins Biopython to a reviewed version.

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
SKILL.md:8
Finding

Unpinned Third-Party Dependency Installation

Content
View full analysis

Vulnerability Details

File Location: SKILL.md, line 8
Vulnerability Type: Unpinned third-party dependency
Risk Level: Medium

Vulnerable Code

yaml
metadata: {
  "openclaw": {
    "emoji": "🧬",
    "requires": { "bins": ["python3"], "install": ["uv pip install biopython"] }
  }
}

Technical Analysis

The installation command resolves and installs biopython without an exact version constraint or an integrity hash. Consequently, the effective dependency can change after the Skill has been reviewed. Security and behavior depend on whichever package release the configured package index serves at installation time.

This is a supply-chain weakness rather than evidence that the current Biopython package is malicious. Exploitation would require the package distribution channel, package publisher account, resolver configuration, or selected future release to be compromised or otherwise untrusted.

Attack Path

  1. An attacker compromises the relevant package publication channel, publisher account, configured package index, or dependency-resolution environment.
  2. The attacker makes a malicious or compromised version of biopython available for resolution.
  3. During Skill setup, uv pip install biopython selects that mutable version because no reviewed version or hash is enforced.
  4. Package installation or a later from Bio import Entrez import executes attacker-controlled package code.
  5. The payload operates with the privileges and environmental access of the user or Agent process performing the installation or running the Skill.

Impact Assessment

Successful exploitation could permit arbitrary code execution under the installing or executing account. The resulting scope may include reading or modifying files accessible to that account, accessing environment variables and process credentials, making network requests, tampering with the Skill runtime, and affecting generated res ...[truncated 175 chars]

Remediation
View remediation

Remediation Suggestions

  • Pin Biopython to a reviewed exact version rather than resolving the latest available release, for example:

    yaml
    "install": ["uv pip install biopython==<reviewed-version>"]
    
  • Generate and maintain a lockfile that records all transitive dependency versions.

  • Enforce cryptographic hashes for downloaded distributions, such as through a requirements file used with --require-hashes.

  • Restrict installation to the expected authenticated package index and prevent untrusted index or mirror configuration.

  • Prefer reviewed wheel artifacts and verify their provenance and integrity in CI before release.

  • Run dependency installation and Skill execution in an isolated, least-privileged environment without unnecessary credentials or sensitive filesystem access.

  • Add automated dependency vulnerability scanning and a controlled process for reviewing and updating pinned versions.

Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (5)

Natural-Language Policy Violations

High
Category
Not specified by scanner
Confidence
98% confidence
Finding

The instructions require translating every article title into Chinese immediately and present this as mandatory, without user opt-in or any documented region-specific justification. This is a natural-language policy violation because it forces a specific language/locale choice on all users.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

The manifest description describes the skill's function but does not specify when it should be invoked, what user requests should trigger it, or any exclusion conditions. For a markdown/manifest file, this broad description can cause unintended activation because there is no explicit trigger scope or negative examples.

Content

No source excerpt is available for this finding.

Autonomous Decision Making

Medium
Category
Excessive Agency
Confidence
80% confidence
Finding

Skill enables autonomous high-impact decisions without human-in-the-loop verification. Critical operations (destructive commands, financial transactions, data deletion) should require explicit user confirmation.

Content

Scanner excerpt · SKILL.md (reported line 23)May include surrounding context.

md
- The script now supports ISSN lookup, so it is highly accurate.

2. **Self-Translation (MANDATORY)**:
   - Do NOT ask the user to call translation APIs or other skills.
   - You have native-level translation capabilities. Use them.
   - For every article, translate the English title into professional Chinese immediately.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The skill instructs the agent to automatically write a full report to the user's Desktop when more than 20 articles are found, without explicit user confirmation or any privacy warning. Writing files to a prominent local directory can leak sensitive research interests, overwrite existing files, or create persistence of data the user did not ask to store.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The script hard-codes a personal email address and sends it to NCBI Entrez on every request, exposing personal contact information to an external service without user disclosure or consent. This creates a privacy and maintainability issue: the address may belong to an unintended individual, cannot be rotated easily, and may cause attribution, abuse complaints, or account-related issues to be tied to that person.

Content

No source excerpt is available for this finding.