piggy-automation

Overview of piggy-automation skill

What piggy-automation is for

piggy-automation is a Claude skill for automating Piggy tasks through Composio’s Rube MCP server. Instead of hard-coding Piggy API calls, the skill instructs the agent to discover the current Piggy tool schemas at runtime, verify the user’s Piggy connection, and then execute the requested workflow through Rube.

This is best for users who already use Claude-compatible MCP clients and want a safer, schema-aware way to run Piggy operations without manually browsing Composio tool documentation before every task.

Best-fit users and workflows

The piggy-automation skill fits teams or operators who need repeatable Piggy actions inside a broader AI workflow, especially when the exact available tools or fields may change. It is useful when you want the agent to:

• discover which Piggy tools are currently available;
• confirm the Piggy connection is active before taking action;
• map a plain-English task into a Rube MCP execution plan;
• avoid stale assumptions about input fields or tool names.

It is not a standalone Piggy client. It depends on Rube MCP and an authenticated Piggy connection.

Key differentiator: search tools first

The most important behavior in this skill is its “always search first” pattern. The agent should call RUBE_SEARCH_TOOLS before executing a Piggy workflow, using your specific task as the search query. That gives the agent current tool slugs, schemas, recommended plans, and known pitfalls, which is more reliable than asking it to guess from memory.

Adoption constraints to know

Before installing piggy-automation, confirm that your client supports MCP tools and can connect to https://rube.app/mcp. You also need an active Piggy connection managed through RUBE_MANAGE_CONNECTIONS with toolkit piggy. If your environment cannot expose MCP tools to Claude, this skill will not be useful by itself.

How to Use piggy-automation skill

piggy-automation install and setup path

Install the skill in a compatible skills environment, for example:

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

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

https://rube.app/mcp

After installation, open composio-skills/piggy-automation/SKILL.md first. This repository has a compact structure with the skill instructions concentrated in SKILL.md, so there are no separate rules/, scripts/, or resources/ folders to inspect.

Required connection checks before running tasks

A good piggy-automation usage flow starts with tool and connection validation:

1. Confirm RUBE_SEARCH_TOOLS is available.
2. Use RUBE_MANAGE_CONNECTIONS with toolkit piggy.
3. If Piggy is not ACTIVE, follow the returned authorization link.
4. Re-check the connection status before executing any workflow.
5. Call RUBE_SEARCH_TOOLS for the specific Piggy task you want done.

This order matters. If you skip discovery, the agent may use outdated field names. If you skip connection checks, the workflow can fail after the plan is already built.

Turning a rough goal into a strong prompt

A weak prompt is: “Use Piggy to do this task.”

A stronger prompt for the piggy-automation skill is:

Use the piggy-automation skill. First call RUBE_SEARCH_TOOLS for the specific Piggy operation I describe, then check my Piggy connection with RUBE_MANAGE_CONNECTIONS. If the connection is active, propose the execution plan and required fields before running it. My task is: [describe the Piggy action, target object, filters, dates, account/store/customer context, and expected output].

Include the action, the object you care about, any constraints, and what you want returned. For example, specify whether you want a summary, a created/updated record, a lookup, or a confirmation before execution.

Practical workflow for reliable results

Use piggy-automation in two phases when the task affects real data:

1. Discovery and plan: Ask the agent to search Piggy tools, inspect schemas, identify required fields, and show the proposed call.
2. Execution: Approve the plan only after required inputs and side effects are clear.

This pattern is especially important for update, creation, redemption, or account-related operations where a wrong field or target could cause unwanted changes.

piggy-automation skill FAQ

Is piggy-automation only for Claude?

The skill is written for Claude-style skill execution and MCP tool use, but the underlying workflow depends on Rube MCP and Composio’s Piggy toolkit. If another agent environment can load the skill instructions and call the same MCP tools, the pattern can be adapted. The easiest path is a Claude-compatible client with skills and MCP support.

How is it better than an ordinary prompt?

An ordinary prompt may ask the model to “use Piggy,” but it may not force tool discovery, connection validation, or schema checking. The piggy-automation skill encodes those steps as the default operating procedure. That reduces guesswork and makes the agent more likely to use the current Piggy toolkit correctly.

What should I verify before using it?

Before relying on piggy-automation for Workflow Automation, verify:

• RUBE_SEARCH_TOOLS responds in your client;
• RUBE_MANAGE_CONNECTIONS can list or manage the Piggy toolkit;
• the Piggy connection status is ACTIVE;
• the searched tool schema matches your intended task;
• you understand whether the operation is read-only or changes data.

When should I not use piggy-automation?

Do not use it if you only need static documentation, if your client cannot connect to MCP servers, or if your Piggy task requires custom business logic that should be reviewed by a developer first. Also avoid direct execution for sensitive operations until the agent has shown the discovered tool, required inputs, and expected side effects.

How to Improve piggy-automation skill

Give piggy-automation better task inputs

The skill performs best when your prompt contains operational detail, not just intent. Provide:

• the exact Piggy task you want;
• relevant identifiers, names, dates, filters, or account context;
• whether the task is lookup, creation, update, deletion, or reporting;
• required output format;
• whether the agent should ask before executing.

This helps RUBE_SEARCH_TOOLS return more relevant tool matches and prevents the agent from selecting an overly broad operation.

Common failure modes to prevent

The most common failure is skipping schema discovery and relying on assumed tool fields. Another is trying to execute before the Piggy connection is active. A third is giving a vague goal that makes the tool search too generic.

Prevent these by explicitly instructing the agent to: search for the exact use case, check connection status, list required fields, and ask for missing values before calling an execution tool.

Iterate after the first output

After the first tool search, refine the workflow instead of rushing to execution. Ask:

• Which Piggy tool slug is recommended and why?
• What fields are required versus optional?
• What side effects will this call have?
• What validation can be done before execution?
• Can the task be run as a read-only lookup first?

This turns piggy-automation from a simple connector into a safer workflow automation assistant.

Improve the skill instructions for your team

If you maintain an internal version, add examples for your most common Piggy workflows, approved naming conventions, required review steps, and “never execute without confirmation” rules for sensitive actions. Keep the core requirement intact: always call RUBE_SEARCH_TOOLS first so the skill continues to follow current Composio Piggy schemas rather than stale assumptions.