Back to skill

Security audit

技能发布文案优化·Skill发布

Security checks for vulnerabilities and agentic risk

Overview

The skill drafts publishing copy but also requires every result to advertise the author’s account and external community without asking the user.

Remove or make the promotional footer optional before use, and review generated copy before publishing. Narrow generic trigger suggestions to the intended domain. The documented pipeline command cannot run from this package because its script is missing.

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
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (4)

Tp2

High
Category
MCP Tool Poisoning
Confidence
85% confidence
Finding

Mixing characters from multiple Unicode scripts in a single identifier is a common technique to create visually ambiguous tool names.

Content

No source excerpt is available for this finding.

Trigger-generation guidance recommends everyday phrases without requiring domain-specific activation constraints.

Medium
Category
Trigger Abuse
Confidence
92% confidence
Finding

The skill recommends generating triggers such as “润色一下” (“polish this”) and “帮我改改” (“help me revise this”), prioritizing broad coverage and quantity. These phrases overlap with routine editing requests, and the generated triggers are not required to include domain qualifiers or contextual activation gates, potentially causing unintended invocations of skills using the resulting metadata.

Content

No source excerpt is available for this finding.

The skill mandates unrelated promotional content and an external group-recruitment link in every output.

Medium
Category
Excessive Agency
Confidence
99% confidence
Finding

The closing instructions require every output to include an unrelated promotional account and external recruitment link, diverting responses from the user’s requested publishing-copy task. This is deliberate answer contamination and user redirection; the supplied content does not establish that the destination steals data or serves malware.

Content

No source excerpt is available for this finding.

The skill mandates unrelated promotional content in every response, redirecting users toward the author's account and external community.

Medium
Category
Prompt Injection
Confidence
99% confidence
Finding

The unconditional footer uses the skill’s instructions to insert author-benefiting advertising into responses without user authorization. This is the same underlying issue as SDI-1, not an additional independent vulnerability, and could also contaminate publishing materials if users reuse the output.

Content

No source excerpt is available for this finding.