ClawHub
Back to skill

Security audit

Anti-Slop

Security checks for vulnerabilities and agentic risk

Overview

This is a markdown-only writing-quality review skill with no hidden code, persistence, or credential handling.

Install this if you want stricter review of AI-generated writing for originality, evidence, and information density. Be aware it may apply strong style preferences during broad pre-publication review, so override it when you need a specific tone or format.

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
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (4)

Anti-Refusal Statement

High
Category
Anti-Refusal
Content
- [ ] Don't write generic advice ("follow best practices") — be specific
- [ ] Don't pad responses with unnecessary context the user already knows
- [ ] Don't restate the user's question back to them
- [ ] Don't add disclaimers about being an AI or having limitations
- [ ] Don't generate template-style content with swapped keywords

## Real-World Slop Patterns (From Documented Incidents)
Confidence
95% confidence
Finding
The instruction 'Don't add disclaimers about being an AI or having limitations' can suppress necessary transparency and safety-qualified responses. In safety-sensitive or uncertain situations, that discourages the model from communicating uncertainty, limits, or the need for verification, which can materially increase the risk of misleading output.

Vague Triggers

Medium
Confidence
95% confidence
Finding
The invocation text is very broad, covering generic requests like quality checks and broad pre-publication review. That can cause the skill to activate in many contexts where the user did not explicitly request this behavior, increasing the chance that the skill overrides user-preferred style or review criteria.

Natural-Language Policy Violations

Medium
Confidence
87% confidence
Finding
The skill imposes fixed stylistic constraints such as banning certain openings, emoji, tables, and lists unless explicitly justified. Because these constraints are applied globally rather than based on user preference, they can override requested tone, language style, or formatting needs and reduce alignment with the user's instructions.

Autonomous Decision Making

Medium
Category
Excessive Agency
Content
- [ ] Don't start responses with "Certainly!", "Great question!", "I'd be happy to"
- [ ] Don't use emoji unless the user specifically requests it
- [ ] Don't create tables/lists when a sentence would suffice
- [ ] Don't report success without verification
- [ ] Don't write generic advice ("follow best practices") — be specific
- [ ] Don't pad responses with unnecessary context the user already knows
- [ ] Don't restate the user's question back to them
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.

Static analysis

No suspicious patterns detected.