ClawHub
Back to skill

Security audit

meta-recursive-capability-discovery

Security checks for vulnerabilities and agentic risk

Overview

The skill is a disclosed capability-discovery helper with a small local learner file, but users should know it includes self-evolution language and local persistence.

Before installing, be aware that this skill is framed as a self-improving meta-skill and can record local learner notes if its learner script is used. Review or disable that learner behavior if you do not want cross-session skill state, and require explicit approval before connecting it to any super-agent, memory, or reflection system.

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
89% confidence
Finding
The skill text explicitly describes writing to learner/self-reflection systems and persisting outputs into a continuous self-evolution loop, which implies file or state write behavior despite no declared permissions. Undeclared write capability is dangerous because it hides persistence side effects from reviewers and can enable unauthorized modification of local state, memory, or downstream skill artifacts.

Tp4

High
Category
MCP Tool Poisoning
Confidence
85% confidence
Finding
The skill presents itself as a capability-discovery meta-skill but also claims autonomous self-verification, self-reflection, super-agent orchestration, learner injection, and continuous self-evolution intended to surpass a teacher skill. That scope expansion increases autonomy and persistence beyond a narrow discovery role, making it easier for the skill to trigger chained actions, modify future behavior, or evade operator expectations.

Static analysis

No suspicious patterns detected.