Back to skill

Security audit

podcast-chat-prep

Security checks for vulnerabilities and agentic risk

Overview

This markdown-only skill is not technically invasive, but it is explicitly framed to help users sound like they listened to podcasts they did not hear.

Install only if you want transcript-based study and note preparation. Do not use it to falsely claim firsthand listening, and only process transcripts you are authorized to use, especially for private recordings or copyrighted podcast material.

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
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (10)

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The language "没听过也能深度聊天" / "证明你真的听过" directly incentivizes deceptive use by helping users appear to have consumed content they did not actually hear. The surrounding skill context makes this more concerning because the entire workflow is designed to mass-produce quotes, inside jokes, and persona cues that increase the credibility of the deception.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

The README explicitly encourages processing third-party podcast transcripts and repurposing them into derivative notes, personas, and publication/chat materials, but it provides no safeguards around consent, copyright, privacy, or terms-of-use. In this context, users may upload non-public transcripts or reuse copyrighted content and personal statements at scale, creating legal and privacy risk that the skill normalizes rather than constrains.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

The trigger phrases are broad enough to match ordinary podcast-related requests, which can cause the skill to activate outside its intended niche. That increases the chance the agent applies a deceptive or manipulative workflow when the user only wanted neutral summarization or note-taking.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

The skill openly optimizes for helping a user appear deeply familiar with content they may not have actually consumed, but it does not warn the user about the deceptive nature or social misuse risk. This omission makes it easier to use the tool for misleading others in social, professional, or editorial contexts.

Content

No source excerpt is available for this finding.

Ssd 4

Medium
Category
Not specified by scanner
Confidence
98% confidence
Finding

The skill describes a stepwise workflow to let users 'deeply chat without having listened' and emphasizes producing convincing details, internal jokes, and viewpoint timelines that create the impression of genuine engagement. That is a deception-enablement pattern: it systematically equips a user to misrepresent their experience and credibility to others.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The activation conditions are ambiguous and lack boundaries, so the skill may be invoked for requests that are only loosely related to podcast help. In this skill, that is more dangerous because the workflow explicitly includes generating social talking points and material for users who may not have consumed the content.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

The entire template and usage instructions are written as mandatory Chinese output formats, with no indication that another language may be used based on user preference. Under the policy, forcing a specific language without opt-in is a natural-language policy violation unless the locale restriction is explicitly justified.

Content

No source excerpt is available for this finding.

Missing User Warnings

Low
Category
Not specified by scanner
Confidence
89% confidence
Finding

The README markets the tool as producing materials so a user can sound as if they really listened, which encourages misrepresentation in social or professional settings. While not a technical exploit, this is a genuine policy/safety issue because the skill is positioned to facilitate deceptive communication rather than transparent summarization.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
93% confidence
Finding

This markdown file uses Chinese throughout for the title, template headings, and usage instructions, which effectively forces a specific language for skill use. The policy allows fixed language only when the skill offers opt-in or clearly documents a justified locale-specific constraint, neither of which appears here.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
92% confidence
Finding

The entire template and usage instructions are written as prescriptive Chinese content, with no indication that users may choose another language or locale. Under the policy rule, forcing a specific language without user opt-in is a natural-language policy concern.

Content

No source excerpt is available for this finding.