ClawHub
Back to skill

Security audit

OpenSpec Dev Flow (CN)

Security checks for vulnerabilities and agentic risk

Overview

This is a markdown workflow skill for planning and carrying out code changes; its broad activation and skip-confirmation option need attention but do not show hidden or malicious behavior.

Install this only if you want OpenSpec planning to appear for many development tasks. Expect it to create openspec/ documents and guide code changes; do not use the skip-confirmation phrases unless you are comfortable letting the agent proceed without the staged review prompts.

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
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • 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
Findings (8)

Vague Triggers

High
Category
Not specified by scanner
Confidence
97% confidence
Finding

The trigger scope is extremely broad: it applies to essentially any request to create, develop, implement, refactor, or fix something. Over-broad activation can cause this workflow skill to intercept many ordinary development requests, leading the agent to perform planning and file operations in contexts where the user did not intend to invoke this specific workflow, amplifying the risk of unnecessary or unintended changes.

Content

No source excerpt is available for this finding.

Autonomous Decision Making

Medium
Category
Excessive Agency
Confidence
85% confidence
Finding

Skill enables autonomous high-impact decisions without human-in-the-loop verification. Critical operations (destructive commands, financial transactions, data deletion) should require explicit user confirmation.

Content

Scanner excerpt · CLAUDE.md (reported line 31)May include surrounding context.

md
### Critical Behavioral Constraints

- **Interactive confirmation required**: Each stage requires user approval before proceeding
- **Never auto-confirm**: Always ask before creating files or proceeding to next stage
- **Exception**: User explicitly says "直接搞" or "不用确认" (skip confirmation)

### Artifact Structure

Autonomous Decision Making

Medium
Category
Excessive Agency
Confidence
85% confidence
Finding

The exception permitting 'skip confirmation' enables the skill to proceed through stages without the normal approval gates. In this repository the skill itself is not executable code, but it is an instruction set for agent behavior, so bypassing confirmations can still lead to unintended autonomous file creation or workflow progression.

Content

Scanner excerpt · CLAUDE.md (reported line 32)May include surrounding context.

md
- **Interactive confirmation required**: Each stage requires user approval before proceeding
- **Never auto-confirm**: Always ask before creating files or proceeding to next stage
- **Exception**: User explicitly says "直接搞" or "不用确认" (skip confirmation)

### Artifact Structure

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

Allowing confirmations to be skipped via specific Chinese phrases creates a language-specific bypass that may not be transparent to all users or reviewers. In a skill that governs staged approvals, any hidden or unevenly documented bypass weakens consent boundaries and can result in the agent proceeding with actions without the user's informed approval.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The trigger conditions are broad enough to activate on routine requests to create, develop, or implement something, which can cause the workflow to take over interactions the user did not intend to route through this skill. In an agent skill context, overbroad activation increases the chance of unwanted file creation, process steering, or confusing behavioral overrides across unrelated tasks.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
80% confidence
Finding

该技能的描述与后续指令整体固定为中文，触发条件和交互示例也默认使用中文，没有说明会根据用户语言偏好切换。对非中文用户而言，这构成了未经用户选择的语言约束，属于自然语言层面的 locale 风险。

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The skill establishes a safety invariant that each stage requires user confirmation before producing artifacts or making changes, then later explicitly allows bypassing that safeguard when the user says 'just do it'. In a skill that creates and modifies files, this weakens consent and change-control boundaries, increasing the chance of unintended repository modifications or abuse via ambiguous prompts.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The skill permits skipping confirmation at the user's request even though its normal behavior includes creating artifacts and later applying file modifications. Without a clear safety warning and enforced approval boundary for write actions, a casual or socially engineered prompt can trigger repository changes without the structured review the workflow otherwise promises.

Content

No source excerpt is available for this finding.

Static analysis

No suspicious patterns detected.