Agently Playbook

• Outcome: Pass
• Latest audit: May 29, 2026 at 8:24 PM UTC
• Version:
  0.1.0

Overview

This is an instruction-only Agently planning skill with broad but disclosed routing guidance and no code, credentials, persistence, or hidden actions.

Install this if you want Agently-centered planning and project-structure guidance for model-powered tools. If you want framework-neutral advice or a different framework, state that explicitly in your prompt so the agent does not default to Agently routing.

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 (4)

Vague Triggers

Medium

• Confidence: 93% confidence
• Finding: The skill description and guidance use very broad trigger language such as generic requests to 'build an assistant' or 'design an internal tool,' which can match many unrelated user intents. This creates an over-selection risk where the skill may be invoked when a more specific or safer skill should handle the request, leading to misrouting, reduced policy precision, and potentially unsafe downstream behavior if this playbook steers execution before the owner layer is actually clear.

Vague Triggers

Medium

• Confidence: 92% confidence
• Finding: The file explicitly says the request can begin from a generic scenario without mentioning Agently, which broadens when this skill may be invoked. In a routing or skill-selection system, overly broad entry criteria can cause misrouting, making the agent apply this playbook to requests that need a narrower or safer skill, increasing the chance of incorrect actions or unsafe downstream guidance.

Vague Triggers

Medium

• Confidence: 88% confidence
• Finding: The example phrase "help me kick off a model-powered internal tool" is common, high-level language that overlaps with many ordinary requests. Because it is not specific to this skill's scope, it can trigger this skill for broadly phrased requests that may belong to other capabilities, leading to scope confusion and unsafe or low-quality routing decisions.

Vague Triggers

Medium

• Confidence: 95% confidence
• Finding: The catch-all guidance to "stay here first" for unresolved business, product, or refactor requests is ambiguous and excessively broad. In practice, this creates a default sink that can absorb many unrelated requests, reducing routing precision and potentially bypassing specialized skills with tighter constraints or safer handling for sensitive domains.

VirusTotal

66/66 vendors flagged this skill as clean.