hermes-dangerous-command-approval-patterns - Dangerous Command Approval for Agent Terminals
Regex patterns and approval flow for blocking agent-issued destructive shell commands.
Tags
Updated: 2026-09-16Capabilities
Typical Inputs
Typical Outputs
What this skill does
- Match destructive shell command patterns
- Detect sensitive path variable expansions
- Detect process kill command substitutions
- Block gateway self-termination commands
- Manage per-session approval states
- Evaluate risk using auxiliary LLM
- Persist user-approved allowlist patterns
Inputs
- Agent shell commands
- Session context keys
- Auxiliary LLM
- Configuration file (config.yaml)
Outputs
- Command approval or block decisions
- User approval prompts
- Updated allowlist in config.yaml
Requirements
- Agent shell execution capability
- Python contextvars support
- Auxiliary LLM access