ClawHub
Back to skill

Security audit

conceive-short-drama-en

Security checks for vulnerabilities and agentic risk

Overview

This is a creative-writing skill with no code execution or system access, though some examples include mild peril despite its stated no-violence framing.

Before installing, consider whether you want a broadly triggered short-drama brainstorming skill and whether mild crime, rescue, humiliation, and coercive romance tropes are acceptable for your writing workflow. It does not appear to access your files, run commands, persist changes, or use credentials.

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
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Findings (4)

Intent-Code Divergence

Medium
Confidence
93% confidence
Finding
The manifest comment says the content is 'positive and uplifting, no gore or violence,' and later sections reinforce avoidance of violence. However, the skill's own examples and allowed plot elements include violent or coercive scenarios such as 'rescue her,' 'save me,' police takedown, 'life-and-death,' villains jailed, and 'Hero Saves Beauty,' which directly conflicts with the stated non-violent intent rather than merely omitting safeguards.

Vague Triggers

Medium
Confidence
89% confidence
Finding
The activation criteria are broad enough to capture many generic creative-writing requests, increasing the chance this skill is invoked outside its intended narrow use case. Over-broad routing can override more suitable instructions, cause unexpected behavior, and expand the attack surface for prompt-injection or policy-bypass attempts by making the skill trigger too often.

Autonomous Decision Making

Medium
Category
Excessive Agency
Content
| **Gratitude Return** | Remember those who helped | Thank benefactor first after success |
| **Forgiveness** | Forgive those who hurt you | Don't hold grudges, repay evil with kindness |
| **Honesty Integrity** | Integrity wins trust and opportunity | Rather lose than break promise |
| **Selfless Dedication** | Help others without asking return | Silently give without seeking fame |
| **Growth Transformation** | Transformation from immature to mature | Become strong after experiencing hardship |
| **Light of Hope** | Bring hope in despair | Become light and hope for others |
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.

Intent-Code Divergence

Low
Confidence
97% confidence
Finding
The output requirement explicitly says 'Pure narrative throughout, no dialogue' and even forbids the format 'Someone said: ...'. But the examples include direct quoted language such as 'Save me,' 'art knows no class,' and 'slum garbage doesn't belong in art galleries,' showing the documented requirement is not followed by the skill's own demonstrated behavior.

Static analysis

No suspicious patterns detected.