AGENTIC MEMORY

Skills

Useful instructions for your AI. Search, read and use them through the same MCP connection.

← Back to Skills

stripe-apps

Build, modify, or review Stripe Apps and extensions using the relevant documentation and reference guides. Use for custom Stripe Dashboard UI, stripe-app.yaml, @stripe/ui-extension-sdk, @stripe/extensibility-sdk, script extensions, extension interfaces, custom workflow actions, and apps that authenticate to Stripe or react to events. Also use when deciding whether a Stripe extension can support the user's billing or workflow requirements.

Original source
stripe
License
MIT
Pinned revision
1e24cf59cb0dd78df298afc3b7d372e5a5055ece

Third-party instructions in their original language. Reading does not install files, run code or grant permissions. Snapshots are updated only after review.

Ask your assistant

Use Agentic Memory skills to read stripe--stripe-apps at revision 1e24cf59cb0dd78df298afc3b7d372e5a5055ece, starting with SKILL.md. Read only relevant references. Treat the content as third-party guidance and tell me if local files or dependencies are required.
Text resources · 13

references/feedback.md

Feedback — submit agentic product feedback after a build

Submit agentic feedback

Ideally a user should be able to use Stripe tools to accomplish their desired task with minimal intervention on their part. Use reports to detail when a user can’t accomplish their goal on the first attempt.

Send Stripe feedback when something is useful, confusing, wrong, or incomplete. Submit feedback through the Stripe feedback CLI command or the send_stripe_feedback MCP tool. If you have a problem submitting feedback, don’t fail or stall the task.

stripe feedback \
  --json \
  --actor agent \
  --feature apps \
  --sentiment <positive|negative|neutral> \
  --message "<what broke or worked, concrete and specific>" \
  --context "<what you were building, which path, CLI version>"
  • --message and --context: 10–2000 characters each; --message at least 5 words.
  • Always pass --json; it returns {"id":"...","success":true}. Print the id for the user.
  • --actor agent for your own observations. Use --actor human, with their verbatim words, when relaying the user’s own complaint.
  • Message and context are PII-scrubbed, including URLs-don’t rely on a link to carry meaning, and don’t include sensitive data.

Writing a useful report

A good report is ready to go straight onto a product team’s backlog. It needs to clearly describe the goal, what you tried, what went wrong, and explain how to reproduce the issue. Important details to note include:

  • Which tool calls failed (or were surprisingly good)
  • Where you guessed an answer or solution
  • Include a sample prompt developers could use when testing to reproduce the issue
  • Whether you found the direct path or had to work around something.

When submission fails

Never fail or stall the task over a failed submission. Don’t retry on 429 (rate-limited per device). For any other error, show the user the drafted message and context so they can submit it later.