Security audit

大佬思想蒸馏框架 dalao v3.5

Security checks for vulnerabilities and agentic risk

Overview

This skill is a Review item because it builds AI personas of real colleagues from workplace chats, private messages, documents, emails, and browser/API access.

Install only if you are prepared to handle coworker and company data carefully. Use it only with explicit authorization from the people and workspace involved, avoid private messages or confidential documents unless clearly permitted, review generated persona files before use, and delete stored knowledge/profile files when no longer needed.

SkillSpector

Vulnerability Patterns

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
Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep

Findings (2)

Ssd 3

High

Confidence: 98% confidence
Finding: The skill explicitly treats private/internal communications such as chat groups, meetings, private messages, and verbatim transcripts as preferred source material for persona simulation. This encourages the collection, inference, and reproduction of sensitive non-public information about real people, creating substantial privacy, confidentiality, and potential data-protection risk if users supply or the system normalizes such material.

Ssd 3

Medium

Confidence: 93% confidence
Finding: The skill directs authors to document real organizational position, relationship networks, and collaboration patterns for identifiable people. In context, this can facilitate profiling of real individuals and reconstruction of internal power structures, which may expose sensitive personal or corporate information and make the simulated outputs more invasive or defamatory.

Static analysis

No suspicious patterns detected.