Back to skill

Security audit

导出与维护 Zread 仓库 Wiki

Security checks for vulnerabilities and agentic risk

Overview

ClawScan marks this skill suspicious because scanners found it can modify its own instructions and maintenance source during a content export; the review is incomplete.

Consider whether you are comfortable installing a skill that scanners report may change its own instructions or source during an export. The scan was partial: 5 of 7 components were scanned, and tool-misuse analysis was degraded for two scripts, so the available evidence is incomplete.

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
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • 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 skill authorizes modifying its own implementation and contributing upstream repairs, beyond exporting Zread Wiki content.

Medium
Category
Excessive Agency
Confidence
88% confidence
Finding

The workflow permits persistent changes to the installed skill or its maintenance source during an export, without requiring separate user approval for those local changes. This expands an export task into modification of instructions used by future sessions; the validation requirements and explicit approval for upstream PRs reduce risk but do not authorize or review local instruction changes.

Content

No source excerpt is available for this finding.

The extraction workflow directs the agent to persistently modify the skill's own maintenance source, beyond exporting Wiki content.

Medium
Category
Excessive Agency
Confidence
90% confidence
Finding

The repair steps direct the agent to locate writable skill source, modify it, and persist reusable rules based on external-page observations, without a separate approval gate. Although the workflow prohibits executing page scripts and requires extraction validation, those checks do not prevent erroneous or adversarially influenced changes from becoming trusted behavior in later runs.

Content

No source excerpt is available for this finding.