Back to skill

Security audit

robot-task-protocol

Security checks for vulnerabilities and agentic risk

Overview

The skill clearly explains its robot-control and payment capabilities and requires user confirmation for consequential actions.

Use a trusted gateway and a wallet with spending limits; discovery and repeated status checks also cost money. Confirm hardware safety and each operator change, and report completion only after verifying the work because it releases payment. Gateway ownership checks and escrow guarantees were not independently tested.

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
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (2)

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
87% confidence
Finding

The description does not disclose the ability to submit task-completion assertions. This is a distinct state-changing operation, especially significant in a system where completion governs escrow release. The request targets the expected gateway, but it does not dispatch work, query status, or manage robot registration. The snippet does not establish whether this request actually releases payment.

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 · SKILL.md (reported line 17)May include surrounding context.

md
emoji: "💧"
    requires:
      bins:
        - curl
        - jq
      env:
        - SPRAAY_GATEWAY_URL