ClawHub
Back to skill

Security audit

strategy-gen

Security checks for vulnerabilities and agentic risk

Overview

This skill generates strategy templates and includes disclosed, local-only scripts for writing an output document and optional usage notes.

Installers should expect this skill to create markdown strategy files and, if the learner commands are used, a local learned_patterns.json containing usage notes, errors, and preferences. Avoid putting confidential strategy details into learner notes unless you are comfortable storing them locally, and choose output paths deliberately.

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 (4)

Lp3

Medium
Category
MCP Least Privilege
Confidence
89% confidence
Finding
The skill advertises shell commands that write output files such as `strategy.md` and appears to support additional persistence via the learner workflow, but no permissions are explicitly declared. Undeclared write capability is risky because users and orchestrators may treat the skill as low-risk while it can still create or modify files in the workspace.

Tp4

High
Category
MCP Tool Poisoning
Confidence
96% confidence
Finding
The declared purpose is strategy generation, but the skill also includes a self-learning system that records usage, failures, notes, preferences, and writes learned state to local storage. This mismatch is dangerous because it expands behavior from one-shot content generation into hidden data collection and persistence, which can capture sensitive user context without clear consent or expectation.

Description-Behavior Mismatch

Medium
Confidence
92% confidence
Finding
This file implements a generic cross-skill telemetry and preference store that persists usage history, errors, notes, and user preferences to disk for any skill, which exceeds the declared purpose of a strategy-generation assistant. Even if not overtly exfiltrating data, this creates unnecessary collection and retention of potentially sensitive user metadata, broadens the skill’s authority, and could enable profiling across unrelated skills.

Context-Inappropriate Capability

Medium
Confidence
95% confidence
Finding
The code persistently records user preferences, capability usage, error types, and free-form notes in learned_patterns.json without any access control, retention policy, or purpose limitation tied to the strategy-planning function. In context, free-form notes and preferences may contain sensitive business goals or internal operational details, so unnecessary persistence increases privacy and data-leak risk.

Static analysis

No suspicious patterns detected.