ClawHub
Back to skill

Security audit

Jobs-System

Security checks for vulnerabilities and agentic risk

Overview

This is mainly a local decision-analysis skill, but it includes under-disclosed local persistence and optional user-profile management commands that users should review before installing.

Install only if you are comfortable with the skill storing local decision records and actor/self-inventory data under its own references directory. Avoid using the user-profile CLI commands unless you understand where that data is stored, and treat outputs as strategic drafts that need independent verification.

SkillSpector

Vulnerability Patterns
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • 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
Findings (5)

Lp3

Medium
Category
MCP Least Privilege
Confidence
83% confidence
Finding
The skill appears to use local file-write/persistence capabilities without declaring them in metadata. Undeclared stateful storage is dangerous because it can retain user inputs, decision traces, or profiles outside the user’s expectation and outside normal permission review, increasing privacy and integrity risk.

Tp4

High
Category
MCP Tool Poisoning
Confidence
94% confidence
Finding
The declared skill behavior is a reasoning workflow, but the actual behavior reportedly includes local persistence, retrieval, profile management, and CLI-style administration features not disclosed in the description. This mismatch is security-relevant because hidden capabilities can bypass user trust assumptions, expand attack surface, and enable collection or reuse of sensitive user data under the guise of a harmless analysis skill.

Description-Behavior Mismatch

Medium
Confidence
91% confidence
Finding
The file claims to be a minimal shell for formatting, persistence, and structural validation of decision artifacts, but the CLI also exposes commands to view, seed, add, import, and confirm a separate user advantage profile store. That scope expansion increases collection and mutation of user-profile data beyond the advertised purpose, creating a privacy and trust boundary issue where users may not expect profile management capabilities to be present or invoked.

Context-Inappropriate Capability

Medium
Confidence
92% confidence
Finding
The CLI permits adding, importing, and confirming user profile data in bulk, which enables persistent mutation of personal inference records unrelated to the declared decision-memo shell functionality. In a skill context, hidden or weakly disclosed profile ingestion is more dangerous because it can silently accumulate sensitive user attributes, increase privacy exposure, and make downstream decisions depend on unverifiable or user-unexpected profile state.

Natural-Language Policy Violations

Low
Confidence
74% confidence
Finding
Mandating zero-hedging language suppresses proper expression of uncertainty and can make speculative or weakly supported conclusions appear authoritative. In a decision-analysis skill, this increases the risk of unsafe overconfidence, especially when outputs may influence business, strategy, or other high-stakes choices.

Static analysis

No suspicious patterns detected.