ClawHub
Back to skill

Security audit

Llm Safe Write

Security checks for vulnerabilities and agentic risk

Overview

This skill is a disclosed file-writing helper that changes how an agent writes large or CJK-containing files, with no evidence of hidden persistence, exfiltration, or privilege escalation.

Install this only if you want your agent to use a more cautious multi-step writing workflow for large or non-ASCII files. Expect it to edit files in smaller chunks and occasionally create temporary placeholders or markers during the write process; review important generated files afterward, as you would with any file-mutating coding skill.

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
  • Rogue AgentSelf-Modification, Session Persistence
  • 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
Findings (3)

Session Persistence

Medium
Category
Rogue Agent
Confidence
60% confidence
Finding

Skill establishes unauthorized persistence across sessions via cron jobs, startup scripts, or state files. Session persistence allows an attacker to maintain access beyond the current interaction.

Content

Scanner excerpt · README.md (reported line 1)May include surrounding context.

md
# llm-safe-write

A prompt/skill that reliably writes large files or files containing CJK/special characters, using an incremental Edit strategy instead of direct Write. Works with any AI coding environment that has Write and Edit tools (opencode, Cursor, Claude Code, etc.).

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The trigger list includes phrases like "write failed" and "file won't write," and then instructs agents to use the skill proactively whenever a file may exceed 50 lines or contain non-ASCII characters. These broad conditions can match many ordinary file-editing situations, increasing the risk of unintended invocation despite the skill also listing some scope constraints.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
98% confidence
Finding

The example string says the AI assistant should "please answer questions in Chinese," which imposes a specific language choice. Because this is presented without any user choice or documented locale justification, it conflicts with the policy against forcing a language or locale without opt-in.

Content

No source excerpt is available for this finding.

Static analysis

No suspicious patterns detected.