Back to skill

Security audit

agent-autopilot

Security checks for vulnerabilities and agentic risk

Overview

This automation skill needs review because it makes poorly documented service routing mandatory and persistent.

Before enabling heartbeats, replace the rayinai/pengui routing with providers you approve and establish what project content they may receive. Limit the agent to the intended project, set spending and stopping limits, and review MEMORY.md and HEARTBEAT.md, including automatic commits.

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
  • Output HandlingUnvalidated Output Injection, Cross-Context Output, Unbounded Output
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (2)

The mandatory rules require Git commit messages in Chinese without user opt-in.

Medium
Category
Output Handling
Confidence
99% confidence
Finding

The template explicitly requires Chinese commit messages, and line 35 requires these rules to be copied into every agent's MEMORY.md. No language choice or justification for this output-language restriction is provided.

Content

No source excerpt is available for this finding.

Coding and documentation tasks are routed through named external interfaces without a data-sharing warning.

Medium
Category
Output Handling
Confidence
88% confidence
Finding

The mandatory tool-routing rules direct coding and documentation work through named interfaces without establishing their trust status or requiring authorization to share project content. In this autonomous, heartbeat-driven workflow, those rules could repeatedly send proprietary code, sensitive documents, or credentials to an unapproved provider; the excerpt does not establish that either interface is malicious or that disclosure has occurred.

Content

No source excerpt is available for this finding.

Static analysis

No suspicious patterns detected.