Back to skill

Security audit

Feifei Companion

Security checks for vulnerabilities and agentic risk

Overview

This tutoring skill is not malware, but it asks to build long-term profiles of children and share learning reports without clear privacy controls.

Review this skill carefully before installing for a child or classroom. It should only be used with explicit student/guardian consent, clear controls for what is stored and shared, per-student isolation, short retention, report previews, verified parent recipients, and an easy way to disable or delete profiling. Remove or ignore the AIGC-detection-evasion feature.

Vulnerability Patterns
  • Agent Memory PoisoningWrites attacker-controlled rules into memory that affect later sessions
  • Insecure Skill Coding PracticesFinds exploitable flaws such as hardcoded secrets or command injection
  • Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
  • 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
Findings (3)

T02 · Agent Memory Poisoning

Warning
Location
SKILL.md:794
Finding

Persistent Memory Poisoning Through Untrusted User Feedback

Content
View full analysis
Remediation
View remediation

T09 · Insecure Skill Coding Practices

Error
Location
SKILL.md:921
Finding

Collection and Long-Term Profiling of Sensitive Student Data Without Defined Privacy Controls

Content
View full analysis
Remediation
View remediation

T09 · Insecure Skill Coding Practices

Warning
Location
SKILL.md:288
Finding

Student Progress Disclosure to Parents Without an Explicit Authorization Check

Content
View full analysis
Remediation
View remediation
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • YARA SignaturesMalware Match, Webshell Match, Cryptominer Match
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Findings (8)

YARA rule 'agent_skill_mcp_tool_poisoning_metadata': MCP/tool metadata poisoning indicators in tool schemas or skill manifests [agent_skills]

High
Category
YARA Match
Confidence
80% confidence
Finding

YARA rule matched a hack tool or exploit indicator (offensive tools, reconnaissance, privilege escalation, or exploit frameworks).

Content

Scanner excerpt · SKILL.md (reported line 6)May include surrounding context.

md
---
name: feifei-companion
version: "4.0.0"
author: "巧未来AI & 学来学去学习社 | Qiaofuture AI & Learn2study Club"
brand: "巧未来AI × 学来学去学习社 联合出品 | Qiaofuture AI × Learn2study Club"
description: "飞飞学伴v4.0 - 三位一体AI学习陪伴系统 | UCL共情学习法+HERMES动态学习+主动AI | 菲菲老师/小菲学姐/浩云学长联合辅导 | FeiFei Companion v4.0 - Trinity AI Learning Companion with UCL Empathic Learning & HERMES Dynamic Learning Loop"
metadata:
  clawdbot:
    emoji: "🎓"
    tags: ["飞飞学伴", "FeiFei Companion", "菲菲老师", "小菲学姐", "浩云学长", "UCL共情学习法", "HERMES", "主动AI", "个性化学习", "K12", "知识网络", "亲子沟通", "成长陪伴", "学习社", "巧未来", "Qiaofuture"]
---

<p align="center">

Ssd 3

High
Category
Not specified by scanner
Confidence
99% confidence
Finding

The student profiling design calls for automatic updates after each interaction using cognitive, emotional, preference, habit, and goal data, including inferred emotional and behavioral characteristics. For minors, this is highly sensitive behavioral profiling and can lead to invasive surveillance, unfair inference, and inappropriate personalization or disclosure to parents without adequate safeguards.

Content

No source excerpt is available for this finding.

Ssd 3

High
Category
Not specified by scanner
Confidence
99% confidence
Finding

The growth portfolio and reporting features prescribe long-term retention of detailed student history, habits, emotions, milestones, and family-facing summaries. This creates a durable dossier on a child, increasing the impact of unauthorized access, function creep, and harmful or inaccurate disclosures to guardians or others.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

This section describes daily automatic review, data updates, and parent progress notification for a child user, but provides no consent, minimization, or privacy-boundary warnings. Because the skill is aimed at minors, silent collection and onward sharing of progress data materially raises privacy and trust risks.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

The skill explicitly advertises a capability to 'lower AIGC detection rate,' which facilitates evasion of academic-integrity or provenance checks and is unrelated to a K12 tutoring tool's legitimate purpose. In the context of a child-focused educational assistant, this increases misuse risk by helping users disguise AI-generated homework or essays as human work.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

The trigger and search keywords are broad, common educational terms, making accidental or overly frequent invocation likely in ordinary conversations. Unintended activation becomes more dangerous here because the skill then performs profiling, reporting, and proactive workflow behaviors on potentially sensitive student interactions.

Content

No source excerpt is available for this finding.

Ssd 3

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

This section directs persistent collection and reuse of user interaction data, preferences, and outcome history in a 'knowledge lake' to optimize future behavior. Persistent storage of child learning data without clear minimization, consent, retention, and access controls creates a substantial privacy risk and expands the consequences of any misuse or disclosure.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
98% confidence
Finding

The skill defines extensive student profiling and long-term tracking across cognitive style, emotional state, habits, goals, and performance, yet omits clear warnings and safeguards for sensitive data handling. In a K12 context, this is especially dangerous because it normalizes covert or underexplained profiling of minors and could enable over-collection, inappropriate inference, or family disclosure of sensitive traits.

Content

No source excerpt is available for this finding.