ClawHub
Back to skill

Security audit

AI Coding Toolkit

Security checks for vulnerabilities and agentic risk

Overview

This is a markdown-only AI coding guide with disclosed workflow advice, though users should review any project rules files it helps create.

Before installing, be aware that this skill may lead your agent to create persistent project guidance files and recommend external AI coding tools or models that can cost money. Review any generated rules files, keep approval prompts enabled for sensitive actions, and confirm model or provider choices before running paid tools.

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
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (5)

External Model or Provider Selection

High
Category
Excessive Agency
Content
- Use `--architect` flag for complex features (planner + implementer)
- `/drop` files you don't need to free context window
- `--map-tokens` to control repo map size
- Run `aider --model claude-sonnet-4-20250514` for best code quality

### Claude Code
Confidence
90% confidence
Finding
Skill selects an external model or provider that may use a different account or billing plan than the operator expects. Undisclosed model switches can cause unexpected cost or quota consumption.

Vague Triggers

Medium
Confidence
92% confidence
Finding
The quick-start instruction 'Set up AI coding for my project' is generic enough that it can plausibly appear in normal user interaction and unintentionally trigger this skill. Because the skill influences project setup and agent behavior, accidental invocation could cause the agent to apply unsolicited methodology, generate rules files, or alter workflow guidance in contexts where the user did not explicitly request this skill.

Autonomous Decision Making

Medium
Category
Excessive Agency
Content
| **Tool use** | Reads files, runs commands, opens browser — full agent |
| **Transparency** | Shows every action before executing — audit everything |
| **Custom instructions** | Per-project system prompts |
| **Auto-approve** | Configure which actions need approval |

**Cline Pro Tips:**
- Set spending limits to prevent runaway API costs
Confidence
85% confidence
Finding
The skill promotes configuring 'Auto-approve' for agent actions in a tool that can read files, run commands, and open a browser. Even though the surrounding text mentions transparency and auditing, recommending approval bypasses for a full-capability agent can reduce human oversight and enable unsafe command execution, unintended file access, or destructive edits if used incautiously.

Autonomous Decision Making

Medium
Category
Excessive Agency
Content
| Anti-Pattern | Why It Fails | Do This Instead |
|---|---|---|
| **Prompt and pray** | No verification = bugs in production | Always review, always test |
| **Paste the whole codebase** | Overwhelms context, increases cost | Curate relevant files only |
| **Never start new chats** | Stale context → hallucinations | New task = new chat |
| **Trust without reading** | AI generates plausible but wrong code | Read every line |
Confidence
75% 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.

Vague Triggers

Medium
Confidence
95% confidence
Finding
The skill exposes a wide set of generic natural-language trigger phrases such as 'Set up AI coding for [project]' and 'Review this AI output' without defining strict activation boundaries, required inputs, or safety gates. In an agent ecosystem, this can cause the skill to activate on broad user requests and influence workflows unexpectedly, increasing the chance of overbroad assistance, unintended file guidance, or misuse in contexts the author did not explicitly scope.

Static analysis

No suspicious patterns detected.