ClawHub
Back to skill

Security audit

Hunt — Digital Bounty Hunter

Security checks for vulnerabilities and agentic risk

Overview

This skill is a disclosed hackathon-finding assistant that browses public event sites, tracks chosen events, and sets reminders, with no evidence of hidden or malicious behavior.

Install only if you want the agent to browse public hackathon sites, read your local USER.md preferences, maintain ~/.openclaw/workspace/map.md, and create reminder jobs. Use explicit commands for adding, removing, or clearing entries, and review scheduled reminders periodically.

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
  • System Prompt LeakageDirect Leakage, Indirect Extraction, Tool-Based Exfiltration
  • Rogue AgentSelf-Modification, Session Persistence
  • 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 (5)

Direct Prompt Extraction

High
Category
System Prompt Leakage
Content
- If no suitable hackathons found, say so honestly — don't pad results with low-quality ones.
- Cron job names should follow: `hunt-reminder-<normalized-name>` for easy management.

## Telegram Output Rules

When delivering output to Telegram:
- **No markdown tables** — Telegram renders them as garbled code blocks
Confidence
85% confidence
Finding
Skill contains instructions that could directly expose system prompts, internal rules, or hidden instructions to users or external parties.

Vague Triggers

Medium
Confidence
95% confidence
Finding
Using the single-word trigger "hunt" is overly broad and can easily appear in normal conversation, causing the skill to activate unintentionally. In this skill, unintended activation is more concerning because activation can lead to web browsing, file writes to `map.md`, and reminder scheduling, so a casual phrase could trigger side effects beyond simple chat behavior.

Vague Triggers

Medium
Confidence
95% confidence
Finding
The trigger description is overly broad, including phrases like 'anything related to' hackathon searching or digital nomad lifestyle. This can cause the skill to activate on loosely related user messages and perform network access, file writes, or scheduling actions unexpectedly.

Session Persistence

Medium
Category
Rogue Agent
Content
---
name: hunt
description: Digital bounty hunter skill for finding, tracking, and managing online hackathon opportunities. Trigger when the user says "hunt", "find hackathons", "show map", "add [numbers] to map", or anything related to searching for online/free hackathons, managing a hackathon bucket list, or living a digital nomad/bounty hunter lifestyle. Use to browse the web for upcoming online hackathons, present them as numbered options, add selections to a map.md tracker, and schedule reminders.
compatibility: Requires web fetching (web_fetch tool) to browse hackathon listings from public event pages. Requires cron scheduling for 1-day-before reminders. Requires file read/write for map.md persistence. Network access to mlh.com, devfolio.co, devpost.com, lablab.ai. Node.js 18+ for bundled CLI tools.
license: MIT
metadata:
  author: lloyd-c137
Confidence
90% confidence
Finding
The skill explicitly persists data to map.md and creates cron-based reminders, which introduces ongoing storage of user activity and future automated actions. Even if intended, persistence and scheduled execution expand the blast radius of accidental activation, stale data, or unauthorized modifications to tracked events.

Vague Triggers

Medium
Confidence
97% confidence
Finding
The single-word trigger 'hunt' is highly ambiguous in normal conversation and may be invoked unintentionally. Because this skill can browse the web, read user profile data, write map.md, and schedule reminders, accidental activation increases the chance of unintended side effects or privacy exposure.

Static analysis

No suspicious patterns detected.