ClawHub
Back to skill

Security audit

Pharmaclaw Literature Agent

Security checks for vulnerabilities and agentic risk

Overview

This skill is a disclosed literature-search tool that sends search terms to public research APIs, with no evidence of hidden persistence, credential access, destructive behavior, or unrelated data handling.

Install only if you are comfortable sending literature queries and paper identifiers to public third-party research services. Do not use proprietary, unpublished, patient-specific, or otherwise confidential research terms unless that disclosure is acceptable in your environment.

A.I.G

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

SkillSpector

Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • 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
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (7)

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The manifest description contains an unusually broad set of trigger phrases such as generic requests about 'research on', 'studies about', 'recent papers', and 'state of the art'. This can cause the skill to activate for many unrelated or weakly related user prompts, increasing the chance of overbroad routing, accidental invocation, and misuse in contexts where a more specific or safer skill should respond.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
89% confidence
Finding

The code sends user-derived search terms to external services (PubMed and Semantic Scholar) without any notice, consent check, or data-minimization guard. If users include sensitive research topics, proprietary compound names, unpublished targets, or internal investigation details in the query fields, that information is disclosed to third-party services and may be logged or retained outside the operator's control.

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 · scripts/semantic_scholar.py (reported line 25)May include surrounding context.

python
print(json.dumps({"status": "error", "error": "requests not installed"}))
    sys.exit(1)

BASE_URL = "https://api.semanticscholar.org/graph/v1"
TIMEOUT = 15
FIELDS = "title,authors,year,abstract,citationCount,influentialCitationCount,publicationTypes,journal,externalIds,tldr,url,openAccessPdf"

Missing User Warnings

Low
Category
Not specified by scanner
Confidence
89% confidence
Finding

This entrypoint forwards user-derived query terms to multiple external services (PubMed, Semantic Scholar, ClinicalTrials.gov, bioRxiv) without any disclosure, consent, or apparent minimization at this layer. While this is expected for a literature-search tool and not inherently malicious, sensitive compound names, targets, disease areas, or unpublished research topics could be exposed to third parties, creating a privacy and confidentiality risk.

Content

No source excerpt is available for this finding.

Missing User Warnings

Low
Category
Not specified by scanner
Confidence
92% confidence
Finding

The script sends the user-provided search query directly to the external NCBI PubMed API, which can expose sensitive research terms, patient-related keywords, or other confidential inputs to a third party. In this skill's context, calling PubMed is expected functionality, but the absence of any user-facing disclosure or consent still creates a real privacy/data-handling risk rather than a code-execution flaw.

Content

No source excerpt is available for this finding.

Missing User Warnings

Low
Category
Not specified by scanner
Confidence
87% confidence
Finding

This code file sends user-provided search queries to an external HTTP API, which is a data-transmitting network operation covered by the warning rule. While the module docstring explains the tool's purpose, it does not clearly disclose that user input will be sent to a third-party service, and there is no confirmation prompt or runtime notice before transmission.

Content

No source excerpt is available for this finding.

Missing User Warnings

Low
Category
Not specified by scanner
Confidence
84% confidence
Finding

The function performs an outbound HTTP request containing the provided paper ID to Semantic Scholar. For code files, external transmission of user data should have some visible warning or disclosure, and this file does not explicitly tell the user that identifiers will be sent to a third-party API.

Content

No source excerpt is available for this finding.

Static analysis

No suspicious patterns detected.