ClawHub
Back to skill

Security audit

Free LLM Installer

Security checks for vulnerabilities and agentic risk

Overview

The skill coherently installs and tests free LLM endpoints with disclosed local config changes and outbound provider requests, but users should not send sensitive data through those free services.

Install only if you want WorkBuddy configured to use these third-party free LLM endpoints. Treat them as external services: do not send secrets, private user data, or production code, use --dry-run before refresh or uninstall when unsure, and remember that --watch makes repeated periodic requests until stopped.

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
  • 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
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Findings (7)

Undeclared Tool Scope

Medium
Category
MCP Least Privilege
Confidence
95% confidence
Finding

The skill instructs the agent to run a local Python script that reads and writes configuration files, creates backups, and makes outbound network requests, but it declares no explicit tool scope or permissions. This creates a capability mismatch: an agent may execute sensitive file and network operations without clear user-visible constraints, increasing the chance of unintended config modification or data exposure.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The trigger list contains broad natural-language phrases like '免费 LLM', '免费模型怎么接入', and '验证免费模型可用吗', which can match ordinary conversation and activate the skill unexpectedly. Because the skill performs installation, verification, monitoring, file writes, and network calls, accidental activation could lead to unplanned system changes or external requests.

Content

No source excerpt is available for this finding.

External Transmission

Medium
Category
Data Exfiltration
Confidence
90% confidence
Finding

The skill directs traffic to third-party endpoints and explicitly promotes sending real test requests, including tool-calling verification, to external services. In context, this means user prompts, model metadata, and possibly sensitive local content used during testing or future use may be transmitted to untrusted free LLM providers, creating confidentiality and compliance risk.

Content

Scanner excerpt · SKILL.md (reported line 37)May include surrounding context.

md
| `nvidia/nemotron-3-ultra-550b-a55b:free` | Kilo Gateway | 1000k | ✅ | ❌ |
| `space-bunny-free` | OpenCode Zen | — | ✅ | ✅ |

- **Kilo Gateway**：`https://api.kilo.ai/api/gateway/v1/chat/completions`
  免注册免 key，18 个免费模型，是首选。
- **OpenCode Zen**：`https://opencode.ai/zen/v1/chat/completions`
  免key，但免费层已收紧为仅 `space-bunny-free` 可从 API 直连

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

The module docstring and user-facing CLI description are written as direct instructions in Chinese, with no indication that other languages are supported or that the user can opt into a locale. Under the policy, forcing a specific language without user choice is a natural-language locale violation.

Content

No source excerpt is available for this finding.

External Transmission

Medium
Category
Data Exfiltration
Confidence
60% confidence
Finding

Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

Content

Scanner excerpt · scripts/install_free_llms.py (reported line 70)May include surrounding context.

python
# 内置配置（全部经实测验证）
# ---------------------------------------------------------------------------

KILO_URL = "https://api.kilo.ai/api/gateway/v1/chat/completions"
ZEN_URL = "https://opencode.ai/zen/v1/chat/completions"

# Kilo 网关前面有 Cloudflare，按 User-Agent 拦截，必须带浏览器 UA

External Transmission

Medium
Category
Data Exfiltration
Confidence
60% confidence
Finding

Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

Content

Scanner excerpt · scripts/install_free_llms.py (reported line 583)May include surrounding context.

python
"id": "deepseek-v4-flash",
            "name": "DeepSeek-V4 Flash",
            "vendor": "DeepSeek",
            "url": "https://api.deepseek.com/chat/completions",
            "apiKey": args.deepseek_key,
            "supportsToolCall": True,
            "supportsImages": True,

Context-Inappropriate Capability

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The script’s --refresh path pulls a remote model list from Kilo and automatically appends new free models without validating that they actually satisfy the skill’s stated safety and compatibility assumptions. This expands trust from a fixed allowlist to a mutable third-party service response, creating a supply-chain style risk where unexpected or downgraded models can be installed and later selected by the user.

Content

No source excerpt is available for this finding.

Static analysis

No suspicious patterns detected.