Skills & IDE Integration

Install the hookstream skill in Claude Code, Cursor, Windsurf, and Copilot CLI.

Skills files teach AI coding assistants about specific APIs and tools. Once the hookstream skill is installed, your assistant can create sources and destinations, configure connections with filters and transforms, send test webhooks, debug delivery failures, and tune retry strategies — all by calling the hookstream API directly from your editor.

The skill file lives at hookstream.io/hookstream.skills.md and covers the full API surface, CLI commands, all 8 destination provider types, and common patterns (signature verification, retry strategies, DLQ, circuit breakers).

What the skill enables

Once installed, you can ask your assistant things like:

Create a hookstream source for Stripe webhooks with signature verification, then wire it to an HTTP destination pointing at my staging API.

Set up a connection that only routes payment_intent.succeeded events and transforms the payload to flatten data.object into top-level fields.

Why are my deliveries failing? Check the circuit breaker status and the last 10 delivery attempts for destination dst_xxx.

Create a DLQ alert that fires when more than 5 events land in the dead letter queue in a 10-minute window, and notify my Slack channel.

The assistant will use the hookstream API (or the /cli CLI) to carry out each request, including creating resources, reading logs, and diagnosing failures.

Install the skill

Create the skills directory

From your project root:

bash
mkdir -p .claude/skills

Download the skill file

bash
curl -o .claude/skills/hookstream.md https://hookstream.io/hookstream.skills.md

The skill is auto-loaded by Claude Code on next launch — no manual registration.

Authenticate the CLI (optional but recommended)

Install and log in once so the skill can drive the CLI non-interactively:

bash
npm install -g /cli
hookstream login

This saves an API key to ~/.config/hookstream/config.json. Claude Code will then be able to run commands like hookstream sources create --json without further auth prompts.

Verify

Ask Claude Code: "Create a test webhook session and send a sample order.created event to it." The assistant should call POST https://hookstream.io/v1/agent/sessions, then POST to the returned URL.

Cursor doesn't have a native skills format yet, so feed it the docs as custom context.

Open Cursor settings

Go to Settings → Rules for AI (or AI → Custom Instructions, depending on version).

Add the docs bundle

Add this as a rule or linked context URL:

text
https://hookstream.io/docs.md

That single file (~180KB) contains the entire hookstream documentation bundle.

Optionally paste the skill file inline

For the most compact, task-focused context, paste the contents of hookstream.skills.md directly into your rules. It's tighter than the full docs bundle and tuned for code generation.

Windsurf uses .windsurfrules at the project root.

Create the rules file

bash
curl -o .windsurfrules https://hookstream.io/hookstream.skills.md

Reload the workspace

Windsurf picks up .windsurfrules on next prompt. Test with: "List the hookstream destination provider types I can use."

GitHub Copilot CLI accepts a system prompt via --system-prompt-file.

Download the skill

bash
curl -o ~/.hookstream-skill.md https://hookstream.io/hookstream.skills.md

Invoke with the prompt file

bash
gh copilot suggest --system-prompt-file ~/.hookstream-skill.md \
  "create a hookstream source for GitHub webhooks"

Alternative: point at the docs bundle

For any tool that accepts a URL or raw markdown as context, use one of:

https://hookstream.io/llms.txt — short summary (~5KB) with links to all doc pages
https://hookstream.io/docs.md — full docs bundle (~180KB), quickstart first
https://hookstream.io/hookstream.skills.md — task-focused skill, tighter than full docs
https://hookstream.io/docs/<slug>.md — any single page as clean markdown

These URLs all return text/markdown and are safe to fetch from any AI tool.

If you're iterating quickly, use the "Copy as context" button on any ApiEndpointBlock in these docs. It copies a clean, AI-ready markdown snippet for that specific endpoint — perfect for dropping into a chat without flooding the context window.

Was this page helpful?

Previous AI & LLM Integration

Next Agent API Reference