ClawHub
Back to skill

Security audit

self-improving-meta

Security checks across malware telemetry and agentic risk

Overview

This skill mainly logs agent-infrastructure problems and its write-capable or hook behaviors are disclosed, opt-in, and scoped to local project files.

Before installing, decide whether you want local .learnings files and optional project-scoped reminders. Enable the PostToolUse error detector only in trusted projects because it reads command output from the hook environment, and review any .learnings entries for secrets or private details before committing them.

SkillSpector

Vulnerability Patterns
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • 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)

Lp3

Medium
Category
MCP Least Privilege
Confidence
84% confidence
Finding
The skill includes explicit network-capable installation paths (`clawdhub install` and `git clone`) but does not declare permissions or clearly gate those actions behind an explicit trust/consent model. In agent environments, undocumented network capability can expand the attack surface by enabling remote code retrieval or supply-chain exposure when the skill is invoked.

Tp4

High
Category
MCP Tool Poisoning
Confidence
92% confidence
Finding
The skill is presented as primarily log-only, but it also documents workflows that can generate new skills, prepare patches, and write files when invoked with `--write` or after approval. That mismatch is dangerous because users or higher-level orchestration may grant it lower trust than warranted, leading to unintended file creation or broader self-improvement actions in shared agent infrastructure.

VirusTotal

64/64 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.