ClawHub
Back to skill

Security audit

ai-writing-assist

Security checks across malware telemetry and agentic risk

Overview

This skill is a disclosed writing-assistance API wrapper that sends user-provided copy prompts and optional file URLs to Flyelep, with no evidence of hidden persistence or unrelated behavior.

Install only if you intend to use Flyelep for writing assistance. Do not send secrets, regulated data, private business material, or sensitive file URLs in prompts, and provide the Flyelep API key only at runtime when you are ready for a request to be sent.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (4)

Context-Inappropriate Capability

Medium
Confidence
84% confidence
Finding
The skill directs the agent to use Shell plus local file creation/deletion for a task that only needs an HTTP API call. This expands the attack surface by introducing unnecessary filesystem and command execution behavior, increasing the chance of command/argument injection, accidental secret exposure in temp files, or unsafe handling of user-controlled content.

Vague Triggers

Medium
Confidence
88% confidence
Finding
The activation scope is very broad—'generate copy, optimize prompts, get creative inspiration' overlaps with common user requests and can cause the skill to trigger in many contexts. Overbroad triggering increases the likelihood that unrelated or sensitive user content is sent to the external API without a narrowly justified need.

Vague Triggers

Medium
Confidence
87% confidence
Finding
The prompt-handling section uses ambiguous criteria such as optimizing descriptions, generating copy, or refining prompts, without concrete boundaries. This can lead to opportunistic invocation and unnecessary forwarding of user prompts or referenced files to a third-party service.

External Transmission

Medium
Category
Data Exfiltration
Content
- **方式 B（无 Write 工具）**：使用 Shell 的 .NET API 创建文件（`Set-Content -Encoding UTF8` 会带 BOM，不可用）
- **macOS/Linux 环境**：
  - bash/zsh 默认使用 UTF-8 编码，可直接内联中文 JSON，无需临时文件。命令中使用 `curl`（无需 `.exe`，无需 `--%`）。
  - **推荐内联写法**：`curl -X POST URL -H "..." -H "..." --data-binary 'JSON单行内容'`，一步完成，无需创建/删除临时文件。
  - 也可使用临时文件方式：`curl --data-binary @payload_temp.json`，与 Windows 写法相同，只需把 `curl.exe --%` 改为 `curl`。
- **清理**：API 返回结果后，务必删除 `payload_temp.json` 临时文件（如使用了临时文件）。
Confidence
91% confidence
Finding
curl -X POST URL -H "..." -H "..." --data-binary 'JSON单行内容'`，一步完成，无需创建/删除临时文件。 - 也可使用临时文件方式：`curl --data-binary @payload_temp.json`，与 Windows 写法相同，只需把 `curl.exe --%` 改为 `curl`。 - **清理**：API 返回结果后，务必

VirusTotal

64/64 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.