Self Improvement For All

• Outcome: Pass
• Latest audit: May 28, 2026, 11:24 AM
• Version:
  1.0.0

Overview

This is a local learning-memory helper that stores and exports user-provided notes, with no evidence of hidden networking or destructive behavior.

Install only if you are comfortable with a local memory file that may persist project details. Do not record credentials, tokens, private customer data, or raw production logs, and review exported JSON before sharing it. Choose export paths carefully to avoid overwriting unrelated files.

SkillSpector

By NVIDIA

Vulnerability Patterns

• Data Exfiltration: External Transmission, Env Variable Harvesting, File System Enumeration
• Trigger Abuse: Overly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
• 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

Findings (4)

Missing User Warnings

Medium

• Confidence: 94% confidence
• Finding: The skill encourages recording arbitrary errors, corrections, and workflow details, then exporting them to JSON for sharing, but it does not warn that these entries may include secrets, personal data, internal URLs, stack traces, tokens, or other sensitive operational context. In a learning/logging tool, users are especially likely to paste raw error messages and debugging data, so omission of a clear data-handling warning materially increases the chance of accidental disclosure through local storage or exported files.

Missing User Warnings

Medium

• Confidence: 94% confidence
• Finding: The agent stores user-supplied learning and error content to local JSON files by default, and those fields can easily contain sensitive prompts, API details, credentials, internal URLs, or private user data. In an agent setting, persistent retention without explicit disclosure, minimization, or controls increases the chance of inadvertent data exposure to other local users, backups, logs, or later exports.

Missing User Warnings

Medium

• Confidence: 92% confidence
• Finding: The export function writes the entire accumulated learning and error history to an arbitrary path, which can include sensitive operational context and prior user inputs. Without warning, path restrictions, or data review, this makes bulk exfiltration or accidental disclosure much easier than the normal on-disk storage alone.

Vague Triggers

Medium

• Confidence: 84% confidence
• Finding: The manifest advertises continuous, real-time learning from errors, feedback, and successful patterns, but it does not define explicit activation boundaries, data scope limits, or approval controls for what gets recorded and later reused. In an agent context, this can enable over-collection of sensitive prompts, outputs, or operational context and can cause unsafe persistence of bad or attacker-injected guidance.

VirusTotal

63/63 vendors flagged this skill as clean.