ClawHub
Back to skill

Security audit

Discuss Before Begin

Security checks for vulnerabilities and agentic risk

Overview

This is a planning-discussion skill that is broad and assertive, but its behavior is disclosed, user-gated, and not destructive.

Install this if you want a strict planning gate before coding or decisions. Be aware it may activate for broad planning language and slow direct execution until you explicitly confirm consensus.

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
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (2)

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The skill advertises very broad trigger phrases such as 'help me explore', 'design a solution', and 'I want to do X', which overlap with common user language and can cause the skill to activate in many unrelated contexts. Overbroad activation is dangerous because it can unexpectedly seize control of conversations, impose its rigid workflow, and interfere with safer or more task-appropriate skills, increasing the chance of misrouting and denial of intended assistant behavior.

Content

No source excerpt is available for this finding.

Autonomous Decision Making

Medium
Category
Excessive Agency
Confidence
80% 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.

Content

Scanner excerpt · SKILL.md (reported line 14)May include surrounding context.

md
1. **No flattery**: If a link in the idea is weak, say so directly. Never brush it off with "that works too". Your goal is to make the idea survive contact with reality, not to make the conversation comfortable.
2. **Dialogue, not interrogation**: The user controls the scope. You steer the pace and stay sharp, but you don't coerce, humiliate, or go in circles.
3. **Separate facts from decisions**: Facts you can verify (files, code, tools, web pages, information a subagent can look up) you verify yourself — never ask the user to do it; but **decisions belong to the user** — you don't decide for them, you only lay each decision clearly in front of the user, along with your recommendation.
4. **Never let ambiguity or conflict slide**: When requirements are unclear, contradictory, or the user is just passively nodding along, you **must ask again** — clarify before continuing. Never design on top of ambiguity. This is a hard rule, not a suggestion.
5. **Confirmation gate**: Unless the user explicitly says "we have consensus", you do not start implementing anything.