Intent Engineering - Make Agents Optimize for What You Actually Want

• Outcome: Pass
• Latest audit: May 28, 2026 at 8:53 AM UTC
• Version:
  1.0.0

Overview

This skill is a coherent workspace-local helper for adding user-defined intent and limited recent memory context to subagent prompts.

Before installing, review INTENT.md and the workspace memory files because selected prior memories and routing history may be shown to subagents. Also inspect any existing lib/correction-tracker.js in the workspace, since the loader will call it if present.

SkillSpector

By NVIDIA

Vulnerability Patterns

• Data Exfiltration: External Transmission, Env Variable Harvesting, File System Enumeration
• Prompt Injection: Instruction Override, Hidden Instructions, Exfiltration Commands
• Privilege Escalation: Excessive Permissions, Sudo/Root Execution, Credential Access
• Supply Chain: Unpinned Dependencies, External Script Fetching, Obfuscated Code
• Excessive Agency: Unrestricted Tool Access, Autonomous Decision Making, Scope Creep

Findings (1)

Missing User Warnings

Medium

• Confidence: 89% confidence
• Finding: This code reads from `memory/episodic` and `routing-decisions.log`, extracts snippets and routing metadata, and prepends them to subagent task descriptions. Although this is the feature's purpose, there is no explicit user disclosure that prior workspace memory and routing history may be transmitted to spawned subagents, which is a privacy-relevant behavior for a context-loading component.

VirusTotal

64/64 vendors flagged this skill as clean.