Feature request: recommended PreToolUse hooks for command validation

Problem

Claude Code executes shell commands that fail frequently — wrong flags, missing binaries, bad syntax, unreachable targets. Each failure poisons the context window and triggers retries, compounding token cost.

We parsed a real Claude Code transcript: 152 command failures in one session. Every retry carried the error forward.

Proposed solution

Support recommended or default PreToolUse hooks that validate commands before execution. A pre-execution gate that returns allow or deny before the command reaches the shell.

Existing implementation

Check already ships as a PreToolUse hook for Claude Code. One config change:

{
  "hooks": {
    "PreToolUse": [{
      "matcher": "Bash|PowerShell",
      "hooks": [{
        "type": "command",
        "command": "npx",
        "args": ["@golproductions/check"]
      }]
    }]
  },
  "env": { "GOL_CLIENT_ID": "your_key" }
}

Results

• Without Check: 152 failed commands in one session
• With Check: 2 failed commands in 6 hours of heavy usage (blog rewrites, deployments, npm publishes, VS Code extension builds)
• Reduction: 98.7%

Sub-100ms validation. $0.0068 AUD per check. Local binary existence checks run instantly with zero API cost.

Why this matters for Claude Code

Every blocked invalid command is a retry that never happens. Every avoided retry is tokens the user keeps. Pre-execution validation makes Claude Code feel more reliable without changing the model.

Links

• Case study: 152 vs 1
• npm package
• Documentation