Security audit

Multi-viewpoint Debates

Security checks across malware telemetry and agentic risk

Overview

This skill appears to be a legitimate debate helper, but users should avoid putting sensitive information into prompts or saved debate archives.

Install only if you are comfortable sending the debate topic and supporting context to multiple sub-agent sessions. Do not include secrets, credentials, customer records, private personal details, regulated data, or sensitive internal URLs unless you have authorization and have redacted them. Review transcript excerpts before saving them, keep archives private, and delete old debate files when they are no longer needed.

SkillSpector

By NVIDIA

Vulnerability Patterns

Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
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 (6)

Missing User Warnings

Medium

Confidence: 88% confidence
Finding: The documented workflow explicitly tells users to save debate outputs and update an index, but it does not warn that the decision prompt and resulting discussion may contain sensitive business, personal, or strategic information that will be persisted to local markdown files. In a skill centered on decision analysis, users are likely to provide confidential material, so silent local retention increases the risk of unintended disclosure through shared directories, backups, version control, or later publication.

Missing User Warnings

Medium

Confidence: 95% confidence
Finding: The skill explicitly encourages using 'actual context,' including specific metrics, real user data, and reference URLs, but provides no minimization or privacy warning. That creates a realistic risk that operators will send sensitive business, customer, or personal data into sub-agent prompts and external tooling unnecessarily.

Missing User Warnings

Medium

Confidence: 86% confidence
Finding: The file explicitly promotes an archive system and pattern analysis over users' decision discussions, but provides no warning about retention, sensitivity, or consent. Decision debates may include business strategy, personal dilemmas, or confidential context, so silently encouraging archival can expose sensitive information or create privacy/compliance issues.

Missing User Warnings

Medium

Confidence: 89% confidence
Finding: The guide tells users to copy full responses from session transcript files stored under a local agent sessions directory, but it provides no warning that those transcripts may contain sensitive prompts, private project details, secrets, or personal data. Because this skill encourages archiving and sharing debate outputs over time, it increases the chance of unnecessary retention and disclosure of sensitive content beyond the original session.

Ssd 3

Medium

Confidence: 96% confidence
Finding: The skill instructs users to collect full persona responses, save them in markdown archives, and maintain an index for later reuse. If prompts or outputs contain sensitive decision context, customer details, internal strategy, or copied transcripts, this creates persistent local disclosure risk and expands the blast radius of any accidental inclusion of sensitive data.

Ssd 3

Medium

Confidence: 97% confidence
Finding: The guidance to include real user data, specific metrics, and reference URLs in debate prompts increases the chance of disclosing confidential information to spawned sub-agents and then preserving it in archives. The danger is amplified because the skill normalizes broad context-sharing without any caution about sensitivity, authorization, or downstream storage.

VirusTotal

65/65 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.