ClawHub
Back to skill

Security audit

summerizeryoutube

Security checks for vulnerabilities and agentic risk

Overview

This skill is a disclosed YouTube summarizer that calls local transcript and retrieval tools, with no artifact evidence of hidden execution, exfiltration, or privilege abuse.

Before installing, confirm you trust the local backend running on localhost:8000, because it retrieves transcripts and may store transcript embeddings in its vector database. The skill itself does not include executable code or request credentials, but its behavior depends on that local service.

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
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • YARA SignaturesMalware Match, Webshell Match, Cryptominer Match
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Findings (4)

YARA rule 'agent_skill_prompt_injection_hidden_instructions': Prompt injection or hidden instructions embedded in AI agent skill text [agent_skills]

High
Category
YARA Match
Content
# YouTube Summarizer & Q&A Assistant

## Overview

This skill turns OpenClaw into a YouTube research assistant.

It enables:
- Structured video summaries
- Context-grounded Q&A
- Multi-language responses (English + Hindi)
- No hallucinations (answers strictly from transcript)

The backend handles:
- Transcript retrieval
- Chunking
- Embeddings
- Vector similarity search (RAG)

This skill handles:
- Reasoning
- Tool orchestration
- Output formatting

---

## Tool Usage Policy (STRICT)

You MUST follow these rules:

### 1️⃣ When user sends a YouTube URL

If the message contains:
- youtu
Confidence
80% confidence
Finding
YARA rule matched a hack tool or exploit indicator (offensive tools, reconnaissance, privilege escalation, or exploit frameworks).

Vague Triggers

Medium
Confidence
91% confidence
Finding
The system prompt uses broad natural-language triggers like 'If user sends YouTube link' and 'If user asks question' without defining tighter conditions, scope boundaries, or validation requirements. This can cause unintended tool invocation on loosely matching inputs, increasing the chance of unnecessary processing, privacy issues, or misuse of backend services.

Natural-Language Policy Violations

Low
Confidence
87% confidence
Finding
The system prompt states 'Support English and Hindi,' which imposes a language/locale constraint without indicating that the user can choose another language or opt in to those languages. The policy requires offering language or locale choice unless the restriction is clearly justified and documented.

Natural-Language Policy Violations

Low
Confidence
88% confidence
Finding
The markdown states 'Default language: English' and only switches to Hindi when the user explicitly requests it. This imposes a language default rather than offering a neutral language choice or explicit opt-in, which can violate language/locale policy requirements.

Static analysis

No suspicious patterns detected.