resolve-mcp: Skill for Claude Code

resolve-mcp Conventions

Tool Module Pattern

Every tool function follows this structure:

from .config import mcp
from .resolve import get_resolve, _boilerplate

@mcp.tool
def resolve_<object>_<action>(param: type) -> str:
    """Docstring describes what the tool does.

    Explains parameters and return value.
    """
    resolve, project, media_pool, timeline = _boilerplate()
    # ... Resolve API calls ...
    return "Success message"

Naming Convention

• All tool functions: resolve_<object>_<action> (e.g., resolve_list_projects, resolve_add_marker_at)
• All tool modules: <category>_tools.py (e.g., project_tools.py, color_tools.py)

Imports

• Always use relative imports: from .config import mcp, from .resolve import get_resolve
• Never use absolute resolve_mcp. imports

Error Handling

• Use @safe_resolve_call decorator from errors.py for tools that call Resolve API
• For tools that don't use _boilerplate(), check get_resolve() returns non-None
• Return error strings (not exceptions) — MCP tools communicate via return values

Gemini AI Tools

• Check if client is None before any Gemini call
• Import from google.genai import types inside the function body, not at module level
• Return a clear error message when Gemini is unavailable

Adding a New Tool Module

1. Create resolve_mcp/<name>_tools.py
2. Add from . import <name>_tools # noqa: F401 to __init__.py
3. Follow the tool module pattern above