Refactored agent commands

Author: Tom Brandenburg

.claude/PRPs/features/completed/add-prp-core-runner-skill.md

@@ -0,0 +1,197 @@
+#!/usr/bin/env -S uv run --script
+"""Invoke a Claude Code slash command from Python.
+
+Usage:
+    uv run .claude/PRPs/scripts/invoke_command.py prp-core-create "Add JWT authentication"
+    uv run .claude/PRPs/scripts/invoke_command.py prp-core-execute my-feature --interactive
+    uv run .claude/PRPs/scripts/invoke_command.py .claude/commands/prp-core/prp-core-pr.md "Add auth feature"
+"""
+
+from __future__ import annotations
+
+import argparse
+import subprocess
+import sys
+from pathlib import Path
+
+ROOT = Path(__file__).resolve().parent.parent.parent.parent  # project root
+
+
+def resolve_command_path(command: str) -> Path:
+    """Resolve command name or path to full .md file path.
+
+    Args:
+        command: Either a command name (e.g., "prp-core-create") or full path
+
+    Returns:
+        Path to the command .md file
+    """
+    # If it's already a path, use it
+    if command.endswith(".md"):
+        path = Path(command)
+        if path.is_absolute():
+            return path
+        return ROOT / path
+
+    # Otherwise, search in .claude/commands/
+    commands_dir = ROOT / ".claude" / "commands"
+
+    # Try common locations
+    search_paths = [
+        commands_dir / f"{command}.md",
+        commands_dir / "prp-core" / f"{command}.md",
+        commands_dir / "prp-commands" / f"{command}.md",
+        commands_dir / "development" / f"{command}.md",
+        commands_dir / "code-quality" / f"{command}.md",
+    ]
+
+    for path in search_paths:
+        if path.exists():
+            return path
+
+    # Fallback: search recursively
+    for md_file in commands_dir.rglob("*.md"):
+        if md_file.stem == command:
+            return md_file
+
+    sys.exit(f"Command not found: {command}")
+
+
+def strip_frontmatter(content: str) -> str:
+    """Remove YAML frontmatter from markdown content.
+
+    Frontmatter is delimited by --- at the start and end.
+    """
+    if content.startswith("---\n"):
+        # Find the closing ---
+        end_marker = content.find("\n---\n", 4)
+        if end_marker != -1:
+            # Return content after frontmatter
+            return content[end_marker + 5:].lstrip()
+    return content
+
+
+def expand_template(template: str, arguments: str) -> str:
+    """Expand template with arguments.
+
+    Supports:
+    - $ARGUMENTS: all arguments as single string
+    - $1, $2, $3, etc.: individual positional arguments
+    """
+    # Strip frontmatter first
+    template = strip_frontmatter(template)
+
+    # Replace $ARGUMENTS with full argument string
+    expanded = template.replace("$ARGUMENTS", arguments)
+
+    # Replace positional arguments
+    args_list = arguments.split()
+    for i, arg in enumerate(args_list, 1):
+        expanded = expanded.replace(f"${i}", arg)
+
+    return expanded
+
+
+def invoke_command(
+    command_path: Path,
+    arguments: str = "",
+    interactive: bool = False,
+    output_format: str = "text",
+    allowed_tools: str = "Edit,Bash,Write,Read,Glob,Grep,TodoWrite,WebFetch,WebSearch,Task",
+) -> None:
+    """Invoke a Claude Code slash command.
+
+    Args:
+        command_path: Path to command .md file
+        arguments: Arguments to pass to command
+        interactive: Run in interactive mode
+        output_format: Output format for headless mode (text, json, stream-json)
+        allowed_tools: Comma-separated list of allowed tools
+    """
+    # Read and expand template
+    template = command_path.read_text()
+    prompt = expand_template(template, arguments)
+
+    # Build command
+    if interactive:
+        # Interactive mode: pipe via stdin
+        cmd = [
+            "claude",
+            "--allowedTools",
+            allowed_tools,
+        ]
+        subprocess.run(cmd, input=prompt.encode(), check=True)
+    else:
+        # Headless mode: use -p flag
+        cmd = [
+            "claude",
+            "-p",
+            prompt,
+            "--allowedTools",
+            allowed_tools,
+            "--output-format",
+            output_format,
+        ]
+        subprocess.run(cmd, check=True)
+
+
+def main() -> None:
+    parser = argparse.ArgumentParser(
+        description="Invoke a Claude Code slash command",
+        epilog="""
+Examples:
+  %(prog)s prp-core-create "Add JWT authentication"
+  %(prog)s prp-core-execute my-feature --interactive
+  %(prog)s .claude/commands/prp-core/prp-core-pr.md "Add auth" --output-format json
+        """,
+        formatter_class=argparse.RawDescriptionHelpFormatter,
+    )
+    parser.add_argument(
+        "command",
+        help="Command name (e.g., 'prp-core-create') or path to .md file",
+    )
+    parser.add_argument(
+        "arguments",
+        nargs="?",
+        default="",
+        help="Arguments to pass to the command",
+    )
+    parser.add_argument(
+        "--interactive",
+        "-i",
+        action="store_true",
+        help="Run in interactive mode",
+    )
+    parser.add_argument(
+        "--output-format",
+        choices=["text", "json", "stream-json"],
+        default="text",
+        help="Output format for headless mode (default: text)",
+    )
+    parser.add_argument(
+        "--allowed-tools",
+        default="Edit,Bash,Write,Read,Glob,Grep,TodoWrite,WebFetch,WebSearch,Task",
+        help="Comma-separated list of allowed tools",
+    )
+
+    args = parser.parse_args()
+
+    # Resolve command path
+    command_path = resolve_command_path(args.command)
+    print(f"Invoking: {command_path.relative_to(ROOT)}", file=sys.stderr)
+    if args.arguments:
+        print(f"Arguments: {args.arguments}", file=sys.stderr)
+    print(file=sys.stderr)
+
+    # Invoke command
+    invoke_command(
+        command_path=command_path,
+        arguments=args.arguments,
+        interactive=args.interactive,
+        output_format=args.output_format,
+        allowed_tools=args.allowed_tools,
+    )
+
+
+if __name__ == "__main__":
+    main()