多Agent团队编排-运营版

• Outcome: Pass
• Latest audit: May 29, 2026 at 8:34 PM UTC
• Version:
  1.0.0

Overview

This is a documentation-only multi-agent workflow playbook with disclosed, purpose-aligned guidance and no install-time code execution.

Before installing, confirm that any shared directories and spawned-agent permissions are scoped to the project and do not expose secrets or unrelated private workspaces. If you automate the cron-style patterns, define exact schedules, triggers, and human approval points for access or credential-related blockers.

SkillSpector

By NVIDIA

Vulnerability Patterns

• Trigger Abuse: Overly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
• Prompt Injection: Instruction Override, Hidden Instructions, Exfiltration Commands
• Data Exfiltration: External Transmission, Env Variable Harvesting, File System Enumeration
• Privilege Escalation: Excessive Permissions, Sudo/Root Execution, Credential Access
• Supply Chain: Unpinned Dependencies, External Script Fetching, Obfuscated Code

Findings (1)

Vague Triggers

Low

• Confidence: 88% confidence
• Finding: The phrase "Every few hours (or on trigger)" is ambiguous because it does not define what the trigger is or what events should and should not start the workflow. In a patterns document, this lack of specificity can lead to overly broad or inconsistent activation of the task-dispatch pattern.

VirusTotal

65/65 vendors flagged this skill as clean.