ClawHub
Back to skill

Security audit

Gentle Reframe Companion

Security checks for vulnerabilities and agentic risk

Overview

This is a disclosed emotional reframing prompt skill with no executable code or system access, but users should not treat it as crisis or clinical mental-health support.

Before installing, understand that this is a prompt-only coping and reframing aid, not therapy, diagnosis, crisis support, or medical advice. Do not rely on it for self-harm thoughts, suicidal thoughts, violence, inability to stay safe, severe panic, chest pain, or other urgent symptoms; seek local emergency help, a crisis line, or a licensed professional in those situations. The publisher should also tighten the anxiety-attack example and avoid inventing other people's motives.

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
  • 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
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (6)

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The README explicitly promotes self-application of the method during anxiety attacks but provides no boundary guidance, crisis limitations, or advice to seek urgent professional help when symptoms are severe. In a mental-health-adjacent skill, this omission can cause users to rely on the tool in situations involving panic, self-harm risk, or acute psychiatric distress where supportive reframing is insufficient or unsafe as a sole response.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

This skill provides mental-health-adjacent emotional support but the main usage guidance does not prominently warn users what to do in crisis situations. In practice, distressed users may present self-harm, suicidality, or acute danger content, and the absence of clear escalation guidance increases the risk that the skill is used where immediate human or emergency intervention is needed.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Medium
Category
Not specified by scanner
Confidence
88% confidence
Finding

The notes section says the skill 'does not handle self-harm, suicide, violence and other crisis scenarios' at L69-L71, but Scenario 3 explicitly teaches a self-application protocol for use 'during anxiety attacks.' While anxiety attacks are not identical to self-harm or violence crises, this is still crisis-adjacent acute-distress guidance that conflicts with the stated boundary of not handling crisis situations.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
74% confidence
Finding

The title and primary description are presented in Chinese, and one example interaction is also exclusively in Chinese, but the README does not state that language is user-selectable or region-specific. Because the file mixes English and Chinese examples without documenting a language-choice policy, it suggests a possible locale constraint without explicit user opt-in.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
76% confidence
Finding

The skill's main instructional content and prompts are written in Chinese, while later examples are in English, but there is no statement that the skill will adapt to the user's preferred language. This can amount to a language-policy issue because the skill appears to impose or assume a language/locale rather than explicitly offering user choice.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Low
Category
Not specified by scanner
Confidence
90% confidence
Finding

The example encourages the model to speculate about a boss's motives or personal circumstances ('maybe they were criticized, under KPI pressure, slept badly') while presenting this as cognitive reframing. In a mental-health support context, this can reinforce inaccurate narratives, distort reality-testing, and lead users to make decisions based on fabricated explanations rather than evidence.

Content

No source excerpt is available for this finding.

Static analysis

No suspicious patterns detected.