ClawHub
Back to skill

Security audit

Safe Long-Run Mode GPT5.4 + Claude

Security checks for vulnerabilities and agentic risk

Overview

This is a procedural long-running task checklist with no executable code, hidden install behavior, data theft, or persistence beyond disclosed checkpoint files.

Install this only if you want the agent to use a more structured workflow for longer tasks: planning phases, writing checkpoint files, considering subagents, and following the skill's GPT-5.4/Claude routing preferences. Review whether those model preferences and extra local checkpoint artifacts fit your workflow, especially for short tasks where the skill may be more process than needed.

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
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (3)

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The front-matter description activates on very broad conditions such as tasks that 'may run for a while,' touch 'multiple files/systems,' or involve APIs/browser automation. That can cause this skill to be invoked for many ordinary tasks, unintentionally changing agent behavior, routing, or tool usage without a tightly bounded need. In a security context, over-broad activation increases the chance that powerful long-run/autonomous patterns are applied where simpler, safer handling would suffice.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
88% confidence
Finding

The instruction 'Use this skill for tasks that may run long, span multiple systems, or risk losing progress if interrupted' is directionally reasonable but ambiguous and lacks scope limits. Without clear thresholds or counterexamples, an orchestrator may over-apply the skill, increasing unnecessary delegation, checkpointing, and external-system interaction patterns that expand operational risk and complexity.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

The checklist prescribes a fixed routing policy such as 'Coding / implementation | GPT-5.4' and 'Strategic recommendations | Claude' without presenting these as optional or user-selectable. This is a natural-language policy concern because it imposes model-selection behavior rather than offering a language/locale-style preference choice or documenting why the constraint is required.

Content

No source excerpt is available for this finding.

Static analysis

No suspicious patterns detected.