openclaw - 💡(How to fix) Fix feat(exec): add description field for intent-based hook approval [1 participants]

Problem

Currently the exec tool only carries a command string. Hook systems like permission-enforcers must use pattern matching on the command text to decide whether to allow or block an operation. This is fragile and doesn't capture the user's intent.

Proposed Solution

Add a description (or intent) field to the exec tool that carries a short natural-language description of what the command is intended to do. Hook systems can then make intelligent decisions based on intent rather than parsing command strings.

Example

{
  "command": "rm -rf /tmp/cache",
  "description": "Delete temporary cache directory"
}

vs. current pattern-matching approach:

{
  "command": "rm -rf /tmp/cache"
  // hook must guess intent from "rm -rf /tmp" pattern
}

Use Cases

1. Permission Enforcer: Block "delete system files" but allow "delete temp cache" based on intent
2. Plan Mode hooks: Ask user "You're about to delete X — confirm?" with human-readable description
3. Audit logs: Make logs readable instead of opaque command strings
4. Safety guardrails: Distinguish between "I want to install a package" vs "I want to reformat your disk"

References

• Claw-Code (OpenClaw fork) has implemented this in their exec tool
• Related skills: hook-system and permission-enforcer in OpenClaw workspace skills

Constraints

• description should be optional to maintain backward compatibility
• Description is informational — not used as the actual command to execute
• Model should populate this automatically when calling exec