Agency Agents

Security checks across malware telemetry and agentic risk

Overview

This is a prompt-only multi-agent skill with no malware evidence, but it overstates what is included and asks for broad autonomous project execution with under-disclosed local persistence.

Install only if you are comfortable with an incomplete MVP-style prompt package. Treat the 61-agent claim cautiously, avoid putting secrets or sensitive customer data into prompts, give explicit checkpoint and file-write instructions before using the orchestrator, and inspect or delete the ~/clawd/agency-agents output folder after use.

SkillSpector

By NVIDIA

Vulnerability Patterns

Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code

Findings (3)

Missing User Warnings

Medium

Confidence: 95% confidence
Finding: The guide explicitly says agent output is automatically saved to a workspace path, but it does not warn users that prompts and generated content may contain sensitive business data, credentials, customer information, or proprietary material. In an agent skill context, silent persistence increases the chance of unintended local data exposure, retention beyond user expectations, and accidental inclusion in backups, sync tools, or shared directories.

Missing User Warnings

Medium

Confidence: 89% confidence
Finding: The skill explicitly instructs the orchestrator to maintain project state, preserve context between agents, and record decisions/progress, but it does not disclose when or where artifacts may be written or what data may be retained. In an agentic system, undocumented persistence can cause unintended storage of sensitive prompts, requirements, internal notes, or generated deliverables, creating privacy, integrity, and operational risks.

Missing User Warnings

Medium

Confidence: 93% confidence
Finding: The skill promises end-to-end autonomous execution with no human intervention, including error handling and workflow progression, without warning about the consequences of unattended actions. This increases the chance of uncontrolled task execution, unintended file changes, excessive resource use, or cascading mistakes across multiple agents before a user can review intermediate outputs.

VirusTotal

63/63 vendors flagged this skill as clean.

View on VirusTotal