ClawHub

simplify-and-harden

Security checks across malware telemetry and agentic risk

Overview

This is mostly a bounded code-review helper, but it also instructs agents to create durable learning records and promote rules into future agent instruction files without a clear approval gate.

Install only if you want an agent to perform automatic post-task cleanup and simple hardening edits. Before enabling the learning loop, require explicit approval for any writes to `.learnings/LEARNINGS.md`, `AGENTS.md`, `CLAUDE.md`, `.github/copilot-instructions.md`, or similar persistent agent-instruction files.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (2)

Context-Inappropriate Capability

Medium
Confidence
95% confidence
Finding
The skill defines a cross-task learning loop that logs findings and promotes them into persistent agent context files, which extends behavior beyond a bounded post-completion review of the current diff. This creates a durable prompt-modification pathway that can amplify mistaken or adversarially influenced findings across future tasks, increasing the blast radius from one review run to many subsequent agent sessions.

Intent-Code Divergence

High
Confidence
98% confidence
Finding
The documentation establishes a hard scope lock that the agent must only touch code modified in the current task, but later instructs writing promotion-ready patterns into agent context/system prompt files. That contradiction undermines the primary safety boundary and can authorize edits to high-leverage control files outside the task diff, enabling persistent behavior changes and policy drift across unrelated work.

VirusTotal

64/64 vendors flagged this skill as clean.

View on VirusTotal