strava-automation
by ComposioHQstrava-automation helps agents run Strava workflows through Composio Rube MCP by discovering current tools, checking the Strava connection, and executing supported actions safely.
Score: 72/100. This is an acceptable directory listing for users who already use or are willing to use Rube MCP with Composio, because it gives agents a clear trigger, setup path, and execution pattern for Strava automation. The score also signals that it is a lightweight orchestration guide rather than a fully self-contained workflow package, so users should expect the agent to rely on live tool discovery for exact schemas and available operations.
- Clear scope and trigger: it is specifically for automating Strava operations through Composio's Strava toolkit via Rube MCP.
- Includes actionable setup prerequisites, including adding https://rube.app/mcp, verifying RUBE_SEARCH_TOOLS, and activating the Strava connection.
- The repeated instruction to call RUBE_SEARCH_TOOLS first should reduce schema drift and help agents use current tool definitions instead of guessing.
- Execution depends on Rube MCP being available and an ACTIVE Strava connection through RUBE_MANAGE_CONNECTIONS; the skill itself provides no standalone scripts or local fallback.
- Workflow guidance is mostly discovery-pattern based, instructing agents to search current tool schemas first rather than documenting concrete Strava tool calls end to end.
Overview of strava-automation skill
What strava-automation does
strava-automation is a Claude skill for running Strava workflows through Composio’s Rube MCP server. It helps an agent discover the current Strava tool schemas, verify the Strava connection, and then execute actions such as reading activity data or automating supported Strava operations without guessing outdated API parameters.
The key point is not that the skill “knows Strava” statically. Its strongest instruction is to call RUBE_SEARCH_TOOLS first, because Composio tool schemas and available Strava actions can change. That makes the strava-automation skill most useful when you want an AI agent to work against live tool definitions rather than a memorized API shape.
Best-fit users and workflows
This skill fits users who already use Claude or another MCP-capable client and want Strava automation inside a broader workflow automation setup. Good use cases include checking recent activities, building training summaries, syncing Strava-derived information into another process, or preparing repeatable activity-management routines.
It is especially relevant for users of strava-automation for Workflow Automation who need the agent to combine tool discovery, authentication checks, and execution planning. It is less useful if you only want a one-off natural-language explanation of Strava data and do not intend to connect Rube MCP.
Main adoption requirement
The main blocker is setup, not prompting. The skill requires the rube MCP server and an active Strava connection managed through Rube. If RUBE_SEARCH_TOOLS is unavailable, or if RUBE_MANAGE_CONNECTIONS cannot activate the Strava toolkit, the skill cannot perform real Strava operations.
The upstream skill is compact and only ships SKILL.md, so there are no helper scripts, examples folder, or local automation utilities to inspect. Your install decision should be based on whether you are comfortable using Rube MCP as the execution layer.
How to Use strava-automation skill
strava-automation install context
Install the skill from the Composio skill collection, then configure Rube MCP in your AI client:
npx skills add ComposioHQ/awesome-claude-skills --skill strava-automation
The skill itself does not include a project-specific installer. After adding it, configure the MCP endpoint:
https://rube.app/mcp
Then confirm the client exposes RUBE_SEARCH_TOOLS. Use RUBE_MANAGE_CONNECTIONS with toolkit strava and complete the returned auth flow if the connection is not ACTIVE. Do not ask the agent to fetch or modify Strava data until the connection status is active.
Inputs the skill needs
A strong strava-automation usage prompt should include:
- The Strava task you want performed
- Whether the agent should read data, write data, or only plan
- The time range, activity type, athlete context, or target records
- Any safety limits, such as “do not create or update anything without confirmation”
- The expected output format, such as a table, summary, JSON, or step-by-step plan
Weak prompt:
Check my Strava.
Better prompt:
Use strava-automation. First call
RUBE_SEARCH_TOOLSfor current Strava activity-reading tools, then verify my Strava connection. If active, retrieve my runs from the last 14 days and summarize distance, moving time, elevation gain, and any unusually hard efforts. Do not create or update Strava records.
This improves results because it gives the agent a discover-check-execute sequence, a task boundary, and a clear output shape.
Recommended workflow pattern
Use the skill in this order:
- Ask the agent to call
RUBE_SEARCH_TOOLSfor the exact Strava use case. - Review the returned tool slugs, schemas, required fields, and pitfalls.
- Ask the agent to check the Strava connection with
RUBE_MANAGE_CONNECTIONS. - If active, execute the selected tool using the discovered schema.
- Have the agent summarize what it did, what data was returned, and any fields it skipped.
This is better than asking for a direct Strava action immediately. The repository explicitly emphasizes tool discovery because Rube returns current schemas and execution plans.
Repository files to read first
Start with:
composio-skills/strava-automation/SKILL.md
There are no visible README.md, metadata.json, scripts/, resources/, or rules/ files in this skill folder. That makes the skill easy to audit, but it also means you should rely on live RUBE_SEARCH_TOOLS responses for operational detail rather than expecting bundled examples.
strava-automation skill FAQ
Is strava-automation useful without Rube MCP?
No. The strava-automation skill depends on Rube MCP and requires RUBE_SEARCH_TOOLS plus Strava connection management. Without that MCP layer, it can still describe a possible workflow, but it cannot reliably discover tools or perform authenticated Strava operations.
How is this better than an ordinary Strava prompt?
A generic prompt may hallucinate Strava API fields or assume outdated endpoint behavior. This skill pushes the agent to discover current Composio Strava tools first, inspect schemas, and execute through Rube. The benefit is safer tool use and less guesswork, especially when automating repeatable workflows.
Is this beginner-friendly?
It is beginner-friendly if you are comfortable adding an MCP server and completing an OAuth-style connection flow. It is not ideal for users who expect a standalone app, browser extension, or no-code dashboard. The skill is a thin operational guide for an AI agent, not a complete Strava automation product.
When should I not use this skill?
Do not use it for unsupported Strava actions, bulk changes you cannot verify, or workflows where you need strict audit controls beyond what your MCP client and Rube session provide. Also avoid it if you only need static training advice; a normal prompt with exported activity data may be simpler.
How to Improve strava-automation skill
Improve strava-automation prompts
The fastest way to improve strava-automation results is to make the intended operation explicit. Include the action type, time range, filters, and confirmation rules.
Better prompt pattern:
Use strava-automation to discover the current Strava tools for reading activities. Check my connection first. Retrieve cycling activities from this month only. Return a compact table with date, name, distance, moving time, elevation, and average speed. If any write-capable tools appear, do not use them.
This helps the agent choose read-only tools, avoid overbroad retrieval, and produce an output you can verify.
Avoid common failure modes
Common problems include skipping tool discovery, assuming a schema, running before the Strava connection is active, and giving the agent an ambiguous goal such as “organize my Strava.” Prevent these by requiring the agent to show the discovered tool name and required input fields before execution.
For write operations, add a confirmation checkpoint:
Before creating, updating, or deleting anything in Strava, show the exact tool, payload, and expected effect, then wait for approval.
Iterate after the first output
After the first run, refine based on returned fields rather than guessing. If the tool returns activity IDs, sport types, timestamps, or pagination details, use those exact fields in the next prompt. Ask the agent to explain missing data clearly: whether the field is unavailable, absent from the selected tool, or filtered out by your request.
For recurring automation, save the final prompt structure that worked: discovery instruction, connection check, selected action, constraints, and output format. That turns the strava-automation skill from a one-off helper into a repeatable Strava workflow component.