splitwise-automation

Overview of splitwise-automation skill

What splitwise-automation does

splitwise-automation is a Claude skill for running Splitwise workflows through Composio’s Rube MCP server. Instead of guessing Splitwise API calls manually, the skill tells the agent to discover the current Rube tool schemas first, verify an active Splitwise connection, and then execute the appropriate Splitwise operation through the available MCP tools.

It is best for users who already work with Claude or another MCP-capable client and want an agent-assisted way to manage Splitwise tasks such as checking balances, creating expenses, updating groups, or coordinating expense workflows without hand-coding against the Splitwise API.

Best-fit users and use cases

The splitwise-automation skill is a good fit if you want Workflow Automation around shared expenses and are comfortable granting an MCP tool access to your Splitwise account through Composio. It is especially useful for recurring household, travel, roommate, team, or event-expense workflows where the agent needs to inspect available Splitwise tools before acting.

It is less useful if you only need a one-off manual Splitwise entry, cannot use MCP tools, or need offline spreadsheet-only expense tracking.

Main differentiator: tool discovery first

The most important design choice in this skill is its “search tools first” rule. Rube MCP tool schemas can change, and Splitwise operations may require specific field names, IDs, or execution plans. The skill instructs the agent to call RUBE_SEARCH_TOOLS before performing work, which reduces brittle prompts and helps the agent use the current Splitwise toolkit rather than stale assumptions.

How to Use splitwise-automation skill

splitwise-automation install and setup requirements

To install the skill from the repository with a compatible skills CLI, use:

npx skills add ComposioHQ/awesome-claude-skills --skill splitwise-automation

Then configure Rube MCP in your AI client by adding the server endpoint:

https://rube.app/mcp

The upstream skill requires Rube MCP and expects RUBE_SEARCH_TOOLS to be available. Before asking for Splitwise automation, use RUBE_MANAGE_CONNECTIONS or the relevant Rube connection tool to connect the splitwise toolkit. Follow the returned authorization link if the connection is not active, and do not run expense-changing workflows until the connection status is ACTIVE.

Inputs the skill needs for reliable execution

A strong splitwise-automation usage prompt should include the action, people or group context, dates, amounts, currency, split method, and whether the agent is allowed to create or modify data. For example:

“Use splitwise-automation to add a dinner expense to my Barcelona trip group. Total is EUR 96.40, paid by me, split equally between Alex, Priya, Marco, and me. First discover the current Splitwise tools, confirm the active connection, show me the planned fields, then create the expense only after I approve.”

This is better than “add my dinner to Splitwise” because the agent can map the task to the discovered schema, ask fewer follow-up questions, and avoid creating an expense with missing payer, participant, or currency details.

Practical workflow for first run

Start by reading composio-skills/splitwise-automation/SKILL.md; it is the only source file in the skill and contains the required workflow pattern. In practice, the agent should:

1. Call RUBE_SEARCH_TOOLS with the specific use case, not a vague generic query.
2. Use the returned schema and recommended execution plan.
3. Check the Splitwise connection status.
4. Resolve group, user, or expense identifiers before making changes.
5. Present a confirmation summary for create, update, or delete actions.
6. Execute through the discovered Rube tool only after the required inputs are clear.

For read-only tasks such as “show what I owe,” confirmation can be lighter. For writes, always ask the agent to preview the exact operation first.

Prompt patterns that improve output quality

Use explicit constraints when money is involved. Good prompts say things like “do not create anything until I confirm,” “ask if a participant cannot be matched,” “use USD,” “split by exact shares,” or “only search expenses from January 2026.” These instructions matter because Splitwise workflows often fail due to ambiguous group names, partial contact names, currency mismatch, or missing split rules.

For multi-step workflows, keep one business outcome per prompt: reconcile a trip, add one batch of expenses, summarize balances, or update a specific expense. Bundling unrelated changes makes it harder for the agent to validate schemas and increases the chance of acting on the wrong entity.

splitwise-automation skill FAQ

Is splitwise-automation beginner-friendly?

Yes, if you already have an MCP-capable client and are willing to connect Splitwise through Rube. The skill itself is short and direct, but the setup depends on external MCP tooling. Beginners should start with read-only prompts such as balance summaries before allowing the agent to create or edit expenses.

How is this better than an ordinary Splitwise prompt?

A normal prompt can describe what you want, but it does not automatically know the live Rube MCP tool schema. The splitwise-automation skill adds operational discipline: discover tools first, check the Splitwise connection, then execute using current schemas. That makes it more reliable for real account operations than a generic “help me with Splitwise” instruction.

What can block adoption?

The main blockers are missing Rube MCP access, an inactive Splitwise connection, unclear account permissions, or ambiguous Splitwise data such as duplicate group names and similar participant names. The repository does not include helper scripts, extra rules, or reference examples, so users should expect to rely on SKILL.md plus live Rube tool discovery.

When should I not use this skill?

Do not use splitwise-automation for financial actions you are not ready to review. Avoid it if you cannot authorize Composio/Rube access, need a fully auditable accounting system, or require custom business logic beyond what the Splitwise toolkit exposes. For sensitive changes, ask the agent to produce a dry-run plan before execution.

How to Improve splitwise-automation skill

Improve splitwise-automation results with clearer context

The fastest way to improve results is to give structured expense context: group name, payer, participants, amount, currency, date, category, split type, and whether the action is read-only or write-enabled. If you know Splitwise group IDs or user IDs, include them. If you do not, instruct the agent to resolve candidates and ask before choosing between similar matches.

Common failure modes to prevent

Common failures include using stale tool assumptions, skipping connection checks, selecting the wrong group, creating expenses without confirmation, or misinterpreting split rules. Counter these by requiring RUBE_SEARCH_TOOLS at the start of each new workflow and asking for a confirmation table before any mutation.

A useful confirmation table includes: operation, group, payer, participants, amount, currency, split method, date, and tool to be called.

Iterate after the first output

After the first response, do not just say “go ahead” unless the plan is complete. Correct missing fields, clarify names, and ask the agent to re-map the operation to the discovered schema. For example: “Use the same plan, but change the split to exact amounts: me 40, Alex 20, Priya 36.40. Reconfirm before creating.”

This keeps the splitwise-automation guide practical: discovery first, clear financial intent second, execution last.