Phy Large Codebase Workflow

Security checks across malware telemetry and agentic risk

Overview

This is a documentation-only workflow guide; its repo-editing and commit examples are disclosed and aligned with large-codebase development, but users should review before letting an agent commit or open a PR.

Install only if you want a workflow guide for large repositories. Before using its automation examples, keep work on a disposable branch, review generated .claude files and codemaps for sensitive content, and require confirmation before commits, pushes, or PR creation.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
Findings (1)

Missing User Warnings

Medium
Confidence
94% confidence
Finding
The skill explicitly instructs the agent to "commit with descriptive message and open a PR" without requiring confirmation before making repository changes or publishing to a remote collaboration surface. In an agent skill, this can lead to unintended state-changing actions, accidental disclosure of sensitive code or metadata, and workflow abuse if followed automatically.

VirusTotal

No VirusTotal findings

View on VirusTotal