ClawHub
Back to skill

Security audit

meta-deep-research

Security checks for vulnerabilities and agentic risk

Overview

This is a research workflow skill that writes expected research outputs and keeps local learning notes, with privacy caveats but no evidence of exfiltration or destructive behavior.

Install only if you are comfortable with the skill creating research files and keeping a local learned_patterns.json history. Keep research outputs in a dedicated folder, do not put secrets or sensitive personal data into preference or note fields, and delete the learning file if you do not want cross-session memory retained.

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 (3)

Lp3

Medium
Category
MCP Least Privilege
Confidence
76% confidence
Finding
The skill explicitly describes writing research artifacts such as outline.yaml, fields.yaml, results/item_N.yaml, and report.md, but it declares no permissions. Undeclared file-write capability is risky because users and orchestrators cannot accurately assess or constrain side effects, and a public skill that writes files can overwrite workspace content or persist data unexpectedly.

Tp4

High
Category
MCP Tool Poisoning
Confidence
89% confidence
Finding
The documented purpose is a research workflow, but the observed behavior includes tracking arbitrary skill usage, persisting user preferences, recording failures, and maintaining a general learned_patterns.json store. This mismatch is dangerous because it hides data collection and persistence beyond the stated research function, reducing informed consent and enabling unintended profiling or retention of sensitive operational metadata.

Missing User Warnings

Medium
Confidence
89% confidence
Finding
The module persists usage history, error notes, and arbitrary user preferences to disk in cleartext JSON without any sensitivity warning, consent mechanism, minimization, or access controls. In a skill environment, notes and preferences can easily contain sensitive user data, and storing them silently increases the risk of local privacy leakage, unintended retention, and later misuse by other local processes or users.

Static analysis

No suspicious patterns detected.