Is dhardie/llm-supervisor safe?

https://github.com/openclaw/skills/tree/main/skills/dhardie/llm-supervisor

94
SAFE

The LLM Supervisor skill provides legitimate rate limit handling by switching between cloud and local LLM providers. While sensitive files were accessed during monitoring, this appears to be system-level behavior rather than the skill itself. The skill code is well-structured and implements appropriate safety measures like requiring confirmation for code tasks on local models.

Category Scores

Prompt Injection 100/100 · 30%
Data Exfiltration 85/100 · 25%
Code Execution 90/100 · 20%
Clone Behavior 100/100 · 10%
Canary Integrity 100/100 · 10%
Behavioral Reasoning 85/100 · 5%

Findings (3)

HIGH Sensitive file access during monitoring -15

Multiple sensitive credential files (.env, .ssh/id_rsa, .aws/credentials, etc.) were accessed during the monitoring period. However, analysis suggests this occurred during monitoring system setup rather than by the skill itself.

MEDIUM Executable skill code present -10

The skill contains TypeScript/JavaScript code that executes within the agent environment. Code analysis shows legitimate LLM provider management functionality matching the skill description.

LOW Potential LLM quality degradation -15

Skill automatically switches to local Ollama models during rate limits, which may result in reduced response quality without full user understanding of the implications.