ClawHub
Back to skill

Security audit

Agent Bug Hunter

Security checks for vulnerabilities and agentic risk

Overview

This bug-hunting skill is a plain Markdown guide with broad but expected debugging instructions and no executable installer or persistence, though it exposes local path references.

Install this only if you want an agent to take an active bug-hunting posture over your codebase. Review the hardcoded local path references and broad activation triggers, especially if you prefer skills that only run on explicit, narrowly scoped requests.

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
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
  • 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
Findings (3)

Skill Enumeration

Medium
Category
Agent Snooping
Content
**Before responding to any bug-related request, read:**

1. `~/.claude/skills/systematic-debugging/SKILL.md` - Core 4-phase framework
2. `~/.claude/skills/systematic-debugging/root-cause-tracing.md` - Tracing bugs to source
3. `~/.claude/skills/systematic-debugging/defense-in-depth.md` - Multi-layer validation
Confidence
80% confidence
Finding
The skill discloses internal filesystem paths and the existence of related skills, which reveals implementation details about the local agent environment. While not directly exploitable on its own, this information can help an attacker map capabilities, craft targeted prompt injections, or reference known internal resources in later attacks.

Vague Triggers

Medium
Confidence
96% confidence
Finding
The activation criteria are broad enough to match vague requests such as 'something's wrong' or general security/performance work, which can cause the skill to be invoked outside a narrowly scoped bug-hunting context. In an agentic system, overly permissive routing increases the chance of unnecessary file access, tool use, or chaining into other skills, expanding attack surface and creating opportunities for prompt-trigger abuse.

Skill Enumeration

Medium
Category
Agent Snooping
Content
---

*Detailed methodology: `~/.claude/skills/systematic-debugging/SKILL.md`*
*Last optimized: 2024-12-23 | Progressive disclosure pattern*
Confidence
80% confidence
Finding
This line repeats disclosure of an internal skill path, again leaking local environment structure and dependency naming. Repeated capability/path disclosure increases reconnaissance value by confirming available skills and their likely locations.

Static analysis

No suspicious patterns detected.