ClawHub
Back to skill

Security audit

Deep Planner

Security checks for vulnerabilities and agentic risk

Overview

This skill is a transparent planning helper that writes task plans to a local `.todolist/` folder, with no evidence of hidden execution, exfiltration, or destructive behavior.

Install this if you want complex tasks broken into visible steps with local progress tracking. Be aware that plans, assumptions, filenames, and progress notes may be saved under `.todolist/`, so avoid using it for prompts containing sensitive details unless that local persistence is acceptable.

A.I.G

Vulnerability Patterns
  • Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
  • Agent Memory PoisoningWrites attacker-controlled rules into memory that affect later sessions
  • Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
  • Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
  • Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs

SkillSpector

Vulnerability Patterns
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • 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 (4)

Vague Triggers

High
Category
Not specified by scanner
Confidence
97% confidence
Finding

The activation criteria are broad enough to trigger on many normal prompts, which can cause the skill to insert itself unexpectedly into unrelated workflows. In context, that matters because this skill initiates persistent state handling, step gating, and delegation behavior, increasing the chance of unintended file persistence, prompt-surface expansion, and user confusion about what actions will be taken.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Medium
Category
Not specified by scanner
Confidence
87% confidence
Finding

The manifest states this skill "does NOT produce the final output" and is "the project manager, not the worker," but L223 also says "Do not perform content generation... inside this skill"—implying content generation is a relevant risk boundary—and the protocol throughout has the skill emitting user-visible plan/status/final wrap-up content. While planning/status text is expected, the documentation mixes a narrow non-worker role with instructions that blur whether the skill may itself generate deliverable content.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

The skill instructs writing task data to .todolist/ without clear user-facing notice or consent, creating a persistence/privacy risk. In this context, the stored plan may contain sensitive prompts, assumptions, filenames, or operational details, and the auto-resume behavior in Step 0 can surface prior task context into later interactions.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

L223 says "Do not perform content generation, file operations, or network requests inside this skill," but multiple earlier sections instruct the skill to scan .todolist/, read existing task files, write new TodoList files, and update them as execution progresses. That is an active contradiction between the documentation of intended behavior and the documented operational steps.

Content

No source excerpt is available for this finding.

Static analysis

No suspicious patterns detected.