agentcmd: Command for Claude Code

Check

Reads the ## Validation section from README.md, runs the validation command, and reports all issues including errors and warnings. Does not make any fixes.

Variables

• $format: $1 (optional) - Output format: "text" or "json" (defaults to "text" if not provided)

Workflow

1. Read README.md

   • Locate the ## Validation section (search for exact header text)
   • Extract the command shown in the code block under this section
   • If no command is found, report error and stop

2. Run Validation

   • Execute the validation command using the Bash tool
   • Capture the full output including errors, warnings, and info messages
   • Do not attempt to fix any issues

3. Parse and Categorize Issues

   • Analyze the output to identify all issues
   • Group issues by type:
     • Errors: Critical issues that prevent builds/tests from passing
     • Warnings: Non-critical issues that should be addressed
     • Info: Informational messages
   • Count issues in each category

4. Report Results

   • Display the validation command that was executed
   • Show total counts by category (e.g., "5 errors, 3 warnings")
   • Display the full validation output
   • Summarize the validation status (pass/fail)
   • If there are issues, suggest running /fix to automatically resolve them

Report

JSON

IMPORTANT: If $format is "json", return raw JSON output (no ```json code fences, no markdown):

{
  "success": true,
  "validation_command": "bun run build && bun test",
  "issues": {
    "errors": 5,
    "warnings": 3,
    "info": 2
  }
}

JSON Field Descriptions:

• success: Always true if check command completed (even if validation failed)
• validation_command: The command that was executed from README
• issues.errors: Number of errors found
• issues.warnings: Number of warnings found
• issues.info: Number of info messages found

Text

Otherwise, provide this human-readable information to the user: