ClawHub
Back to skill

Security audit

Clarity Variant

Security checks for vulnerabilities and agentic risk

Overview

This skill appears to be a straightforward Clarity Protocol lookup tool with disclosed network use and an optional API key.

Install only if you are comfortable querying clarityprotocol.io and, if you set CLARITY_API_KEY, sending that key as an API header to Clarity Protocol. The inspected code does not show hidden persistence, local data collection, or unrelated outbound destinations.

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
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (4)

Tp4

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding
The code chunk is narrowly focused on one endpoint: /variants/{fold_id}/annotations. It supports fetching annotations and filtering them by agent and annotation type, which aligns with the 'agent annotations' portion of the description. However, it does not retrieve general variant details, AI summaries, fold quality metrics, pLDDT scores, protein mutation analysis, or agent findings from a distinct findings capability. Because the declared purpose presents a broader multi-capability skill while the actual code only implements annotations retrieval, the description materially overstates what this code chunk does.

Tp4

High
Category
MCP Tool Poisoning
Confidence
96% confidence
Finding
The implemented script is narrower than the declared description. It does support one declared capability: fetching agent findings by type for a variant. However, it does not retrieve detailed variant information, fold quality, pLDDT scores, protein mutation analysis, AI summary for the variant, or agent annotations. Its primary behavior is limited to calling the /variants/{fold_id}/findings endpoint and formatting the returned findings. This makes the description materially overstate the skill's capabilities relative to the supplied code chunk.

Tp4

High
Category
MCP Tool Poisoning
Confidence
93% confidence
Finding
The implemented code generally matches the core purpose of getting variant details and exposing an AI summary if that data is present in the API response. However, the description explicitly claims capabilities for 'agent findings by type' and 'agent annotations', and the code does not perform any dedicated retrieval, filtering, or presentation of agent findings or annotations. It only requests a single variant object and prints standard fields plus optional ai_summary and notes. There is no evidence of separate handling for agent findings, annotation structures, or type-based findings retrieval. Therefore the description overstates the implemented capabilities, creating a partial but material mismatch.

Lp3

Medium
Category
MCP Least Privilege
Confidence
94% confidence
Finding
The skill explicitly requires internet access and optionally reads the CLARITY_API_KEY environment variable, but it does not declare any tool scope such as permissions or allowed-tools. This creates an under-specified trust boundary: an agent may invoke network and environment access without clear least-privilege constraints, making secret exposure or unintended outbound requests more likely in broader agent runtimes.

Static analysis

No suspicious patterns detected.