From workbench-ynab
Swap-ready protocol for calling the vendored YNAB MCP. Use whenever a workbench-ynab skill, the orchestrator, or pre-approval config needs a YNAB MCP tool name — never hard-code a tool name; resolve it from this protocol's single source of truth so an MCP swap stays a one-file edit.
How this skill is triggered — by the user, by Claude, or both
Slash command
/workbench-ynab:protocolThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
This skill is the swap-ready indirection layer between workbench-ynab's skills
This skill is the swap-ready indirection layer between workbench-ynab's skills
and whatever concrete YNAB MCP backs them. It exists so the rituals reference
logical operations, and the concrete mcp__plugin_workbench-ynab_ynab__*
tool names live in exactly one place.
ynab-tools.md — the single source of truth for the
complete tool-name list, the pre-approval glob, and the orchestrator tools
list. Reference it; do not copy a name into your own skill or config.docs/mcp-capability-map.md.Never hard-code a raw mcp__plugin_workbench-ynab_ynab__ynab_* tool name in a
skill, command, hook, asset, or config. Centralize every reference in
ynab-tools.md so swapping the MCP stays a one-file edit.
bin/check-tool-name-sources.sh enforces this across the whole tree, with a
narrow allowlist for the two docs and the orchestrator agent's tools:
frontmatter (which Claude Code requires to hold literal names). See
docs/mcp-capability-map.md for the full
allowlist and rationale.
npx claudepluginhub mike-bronner/workbench-ynabCreates, edits, and optimizes skills for Claude Code, including drafting, evaluating with test prompts, iterating on performance, and improving skill descriptions for better triggering accuracy.