retently-automation

Overview of retently-automation skill

What retently-automation does

retently-automation is a Claude skill for automating Retently work through Composio’s Rube MCP server. It is designed for tasks such as finding the right Retently tool, checking the Retently connection, and executing operations only after discovering the current tool schema with RUBE_SEARCH_TOOLS.

Best-fit users and workflow fit

This skill fits teams that already use Retently for customer experience, NPS, survey, or feedback workflows and want an AI agent to operate Retently through an MCP tool layer instead of relying on copy-pasted API assumptions. It is especially relevant for users looking for retently-automation for Workflow Automation, because the skill emphasizes a repeatable tool-discovery pattern before execution.

Key differentiator: schema-first execution

The important adoption point is not that the skill “knows Retently.” Its main value is that it tells the agent to search Rube tools first, inspect the live schema, confirm the Retently connection, and then run the matching tool. That reduces failures caused by stale tool names, changed parameters, or incomplete prompts.

What to review before installing

The repository path contains a single main file: composio-skills/retently-automation/SKILL.md. Read that first. There are no bundled scripts, examples, rules, or reference folders, so your decision should be based on whether the Rube MCP dependency and Retently connection model fit your environment.

How to Use retently-automation skill

retently-automation install context

To use the retently-automation skill, your AI client must support skills and MCP servers. The skill itself requires Rube MCP and declares mcp: [rube]. In practice, installation has two parts:

1. Add or install the skill from ComposioHQ/awesome-claude-skills, path composio-skills/retently-automation.
2. Add Rube MCP as a server in your client configuration using https://rube.app/mcp.

After that, confirm the MCP tools are visible. The skill expects RUBE_SEARCH_TOOLS and RUBE_MANAGE_CONNECTIONS to be callable. A retently-automation install is incomplete until the Retently toolkit connection is active.

Inputs the skill needs from you

A weak request like “update Retently” leaves too much ambiguity. Give the agent the business object, action, filters, safety limits, and expected output. Stronger inputs include:

• The Retently task: create, find, update, list, trigger, sync, or audit.
• The target scope: campaign, survey, contact, company, response, audience, or segment if known.
• Identifiers or filters: emails, account IDs, date ranges, campaign names, tags, or status.
• Constraints: dry run first, limit results, do not modify records without confirmation.
• Desired output: summary table, changed-record list, errors, or next-step recommendations.

Example prompt:

Use retently-automation to find the current Rube tools for Retently, confirm the Retently connection is active, then list recent survey responses from the last 14 days. Do not modify anything. Return the tool selected, schema fields used, result count, and any records missing customer email.

Practical retently-automation usage workflow

A reliable retently-automation usage flow is:

1. Ask the agent to call RUBE_SEARCH_TOOLS for the exact Retently use case.
2. Confirm the returned tool slug and required schema fields.
3. Use RUBE_MANAGE_CONNECTIONS for toolkit retently if the connection state is unknown.
4. If the connection is not ACTIVE, complete the returned authorization flow.
5. Run read-only queries before write operations.
6. For writes, ask for a preview plan and require confirmation before execution.

This matters because the skill’s own source warns that current schemas must be discovered first. Do not prompt the agent to guess Retently parameters from memory.

Files to read first

Start with SKILL.md. Focus on these sections:

• Prerequisites for MCP and Retently connection requirements.
• Setup for connection activation.
• Tool Discovery for the required RUBE_SEARCH_TOOLS pattern.
• Core Workflow Pattern for the order of discovery, connection check, and execution.

Because there are no helper scripts or sample workflows in the skill folder, use the upstream Composio Retently toolkit documentation when you need deeper field-level behavior.

retently-automation skill FAQ

Is retently-automation useful without Rube MCP?

No. The skill is specifically built around Rube MCP. If your client cannot call RUBE_SEARCH_TOOLS and RUBE_MANAGE_CONNECTIONS, the skill loses its core execution path. You could still read it as a workflow note, but it will not function as intended.

How is this better than an ordinary Retently prompt?

A normal prompt may ask the model to infer Retently operations from general knowledge. The retently-automation skill forces a safer pattern: discover available tools, inspect live schemas, check connection state, and only then execute. That is more reliable for operational workflows where stale assumptions can create failed calls or unintended changes.

Is this suitable for beginners?

Yes, if the beginner already has access to an MCP-capable client and can complete the Retently authorization step. It is not a full Retently tutorial. New users should begin with read-only tasks such as listing campaigns, checking connection status, or retrieving recent responses before attempting updates.

When should I not use this skill?

Do not use it when you need direct Retently API code, offline analysis, custom ETL scripts, or workflows outside Composio/Rube. Also avoid it for high-risk bulk updates unless you can provide exact filters, require a dry run, and review the proposed tool call before execution.

How to Improve retently-automation skill

Improve retently-automation prompts with concrete scope

The fastest way to improve retently-automation results is to replace broad goals with operational constraints. Instead of:

Sync Retently contacts.

Use:

Use retently-automation to discover the current Retently contact tools, verify the active connection, then prepare a dry-run plan to update contacts with tag enterprise-qbr where last_survey_sent is empty. Show required schema fields and wait for approval before writing.

This gives the agent enough context to choose tools, avoid unsafe writes, and explain its plan.

Common failure modes to prevent

The most common failures are skipped tool discovery, inactive Retently connection, missing identifiers, and write operations without confirmation. Prevent them by adding these instructions to your prompt:

• “Call RUBE_SEARCH_TOOLS first.”
• “Check RUBE_MANAGE_CONNECTIONS for toolkit retently.”
• “Use read-only lookup before any mutation.”
• “If required fields are missing, ask me instead of guessing.”
• “For bulk changes, return a preview and wait.”

Iterate after the first output

After the first response, inspect whether the agent reported the selected tool slug, required fields, connection state, and execution result. If any are missing, ask for a second pass:

Re-run the planning step. Include the discovered Retently tool slug, required input schema, optional filters, connection status, and the exact fields you still need from me.

This turns the skill from a one-shot automation request into a controlled workflow.

What would make the skill stronger

The current skill is concise and operational, but it would be easier to adopt with example prompts, read-only versus write-operation patterns, common Retently use cases, and failure-handling guidance. Adding small examples for listing survey responses, checking campaigns, and safely updating contacts would improve install confidence without changing the schema-first design.