wachete-automation
by ComposioHQwachete-automation is a Claude skill for Wachete workflow automation through Composio Rube MCP. Use it to discover live Wachete tool schemas, verify an active connection, and run supported operations safely.
This skill scores 67/100, which makes it acceptable but limited for directory listing. It gives agents enough trigger and setup guidance to handle Wachete operations through Rube MCP with less guesswork than a generic prompt, but directory users should understand that it is a thin orchestration guide rather than a fully worked Wachete automation playbook.
- Valid skill frontmatter clearly identifies the trigger domain: automating Wachete via Rube MCP with an explicit `requires: mcp: [rube]` dependency.
- Provides practical setup prerequisites: connect Rube MCP, manage the `wachete` toolkit connection, verify ACTIVE status, and use `RUBE_SEARCH_TOOLS` before execution.
- Includes a repeatable operational pattern for discovering tools, checking connection state, and using current schemas instead of hard-coded assumptions.
- No support files, scripts, reference docs, or embedded Wachete tool schemas are included; the agent must rely on live RUBE_SEARCH_TOOLS discovery.
- Workflow guidance is mostly a generic Rube MCP pattern rather than concrete, end-to-end Wachete automations, so users may need domain knowledge of Wachete tasks.
Overview of wachete-automation skill
What wachete-automation does
wachete-automation is a Claude skill for running Wachete-related automation through Composio’s Rube MCP server. It is designed for agents that need to discover current Wachete tool schemas, confirm an authenticated Wachete connection, and then execute Wachete operations through MCP instead of relying on hard-coded API assumptions.
The core value is not a large local codebase; it is a repeatable workflow pattern: search Rube tools first, verify the wachete toolkit connection, execute the selected tool, and handle returned schemas and errors carefully.
Best fit for Workflow Automation users
This wachete-automation skill fits users who already use, or are willing to use, Rube MCP and Composio as the integration layer for Wachete. It is most useful when you want an AI assistant to perform Wachete operations inside a broader workflow automation session, such as monitoring configured web pages, checking watch status, or managing Wachete-related tasks exposed by the live toolkit.
It is less useful if you want a standalone Wachete SDK, a direct REST API wrapper, or a local script that works without MCP.
Key adoption requirements
Before installing or invoking the skill, confirm three things:
- Your client supports MCP tools.
- Rube MCP is configured with
https://rube.app/mcp. - The Wachete toolkit connection can be activated through
RUBE_MANAGE_CONNECTIONS.
The most important operational rule is that the agent should call RUBE_SEARCH_TOOLS first. Wachete tool names, required fields, and schemas may change, so the skill intentionally depends on live tool discovery rather than static examples.
How to Use wachete-automation skill
wachete-automation install and setup path
Install the skill in a Claude skills-compatible environment, for example:
npx skills add ComposioHQ/awesome-claude-skills --skill wachete-automation
Then configure Rube MCP in your client by adding:
https://rube.app/mcp
After MCP is available, verify that RUBE_SEARCH_TOOLS responds. Next, use RUBE_MANAGE_CONNECTIONS with toolkit wachete. If the connection is not ACTIVE, follow the authentication link returned by Rube and re-check the connection before asking the agent to run Wachete actions.
Inputs the skill needs from you
A vague request like “automate Wachete” is not enough. Give the agent the practical task, the target object, and any constraints. Strong inputs include:
- The Wachete action you want, such as checking status, finding available operations, or managing a watch.
- Any page URL, watch identifier, project name, label, or account context you know.
- Whether the action should be read-only or allowed to modify Wachete configuration.
- How you want results summarized, such as a brief status report, a table, or next-step recommendations.
Example prompt:
“Use the wachete-automation skill. First search Rube for current Wachete tools and schemas. Confirm the wachete connection is active. Then find the available tool for checking my Wachete watches and return a concise table of watch name, URL if available, status, and any errors. Do not create or modify watches unless you ask me first.”
Practical wachete-automation usage workflow
A reliable wachete-automation usage pattern is:
- Ask the agent to invoke the skill and search tools for your exact Wachete task.
- Require a connection check before execution.
- Have the agent restate the selected tool slug and required fields.
- Approve any destructive or configuration-changing action.
- Ask for the raw error message if a tool call fails, then retry with corrected fields.
For safer workflow automation, separate discovery from execution. First prompt the agent to list available Wachete operations. Then choose the operation to run. This reduces failed calls caused by guessed parameters.
Repository files to read first
This skill is compact. Start with composio-skills/wachete-automation/SKILL.md; it contains the useful setup and workflow instructions. There are no visible companion scripts/, references/, resources/, or rules/ folders in the provided repository preview, so most behavior depends on Rube MCP tool discovery at runtime rather than local helper files.
The Composio toolkit documentation linked in the skill, https://composio.dev/toolkits/wachete, is the next place to check if you need broader context about the Wachete toolkit.
wachete-automation skill FAQ
Is wachete-automation a Wachete API client?
No. wachete-automation is a skill that guides an AI agent to use Wachete through Rube MCP and Composio tools. It does not provide a direct JavaScript or Python client, and it does not remove the need for an active Wachete connection.
Why is tool discovery required every time?
The skill explicitly says to call RUBE_SEARCH_TOOLS first because live MCP tool schemas are the source of truth. This matters for Wachete automation because a guessed field name or stale tool slug can make the agent fail even when your high-level goal is clear.
Is this suitable for beginners?
Yes, if you can configure MCP and complete the Wachete connection flow. Beginners should keep the first task read-only, such as discovering available Wachete operations or listing current watches, before allowing the agent to create, update, or delete anything.
When should I not use this skill?
Do not use the wachete-automation skill if your environment cannot access MCP tools, if you need offline automation, or if your organization requires direct control over every API request without a Composio/Rube abstraction layer. It is also not the best fit when you need a fully documented local library with typed function signatures.
How to Improve wachete-automation skill
Improve prompts with explicit Wachete intent
The fastest way to improve wachete-automation results is to state the job in operational terms. Instead of “check Wachete,” say:
“Search Rube tools for the current Wachete schema for listing watches. Confirm the Wachete connection is active. If the tool supports filters, use only active watches. Return watch name, monitored URL, last check time, and alert status if those fields exist.”
This tells the agent what to discover, what to verify, and how to shape the output without inventing unsupported fields.
Reduce failures by separating discovery and action
Common failure modes include using a stale schema, assuming a tool exists, skipping connection activation, or running a modifying action before the user has confirmed the target. Avoid this by asking the agent to show the discovered tool slug, required inputs, and planned execution before it calls the tool.
For higher-risk tasks, add: “Do not make changes until I approve the exact tool call and parameters.”
Iterate after the first tool response
If the first response is incomplete, do not restart with a generic prompt. Use the returned schema and error details. Ask the agent to revise only the missing fields, retry with the same session when possible, and explain whether the issue is authentication, missing input, unavailable Wachete data, or a toolkit limitation.
This keeps the workflow grounded in Rube’s actual response instead of letting the agent guess.
What would make the skill stronger
The repository skill could be improved with a few practical additions: example prompts for common Wachete tasks, a read-only safety pattern, troubleshooting notes for inactive connections, and sample output formats. Until those are added upstream, users should compensate by giving clear task intent, requiring RUBE_SEARCH_TOOLS, and approving any Wachete configuration changes before execution.