Ambush Stream Skills

Agent skills for creating, processing, delivering, reviewing, and managing personalized Ambush news streams.

Install in OpenClaw

Ambush is packaged as a native OpenClaw plugin so the skill and its remote MCP server arrive together. It needs OpenClaw >=2026.8.1. Install it with one command:

openclaw plugins install clawhub:@ambush-ai/ambush-streams

The first Ambush tool call returns a sign-in link for that message sender. Each sender connects their own Ambush account; tokens are not shared between chat users. Sign-in links are single-use bearer links, so use this flow only in channels whose participants are mutually trusted. The Gateway needs an externally reachable callback origin:

openclaw config set gateway.publicOrigin https://your-gateway.example.com

OpenClaw then handles OAuth discovery from https://api.ambush.ai/mcp; do not paste tokens into OpenClaw config or chat. Verify the installation with:

openclaw plugins inspect ambush-streams --runtime --json
openclaw skills info manage-ambush-streams

To test unreleased changes, install straight from GitHub:

openclaw plugins install \
  git:github.com/Ambush-AI/ambush-stream-skills@main \
  --force

Shared account for cron and headless runs

Cron, heartbeat, subagent, and other headless runs do not have a trusted message sender to authenticate. On a single-tenant Gateway, replace the plugin's per-requester definition with an operator-managed shared connection:

openclaw mcp set ambush-streams \
  '{"url":"https://api.ambush.ai/mcp","transport":"streamable-http","auth":"oauth","oauth":{"identity":"shared"}}'
openclaw mcp login ambush-streams

This intentionally makes one Ambush account available to every eligible run on that Gateway. Do not use it on a multi-user Gateway. To return to per-requester OAuth, log out and remove the operator override:

openclaw mcp logout ambush-streams
openclaw mcp unset ambush-streams

Install as an agent skill

Install the Ambush stream-management skill with the open skills CLI:

npx skills add Ambush-AI/ambush-stream-skills --skill manage-ambush-streams

To install it globally for Codex without prompts:

npx skills add Ambush-AI/ambush-stream-skills \
  --skill manage-ambush-streams \
  --agent codex \
  --global \
  --yes

The skill declares the production Ambush Streams MCP server at https://api.ambush.ai/mcp. On supported OpenAI surfaces, connect your Ambush account with OAuth when prompted. Other agent hosts may require configuring that remote MCP server separately.

Install in eve

An official eve integration is proposed to install the skill and OAuth MCP connection together with eve add ambush-streams.

Until that registry item and the pending skills.sh re-index land, install the current skill directly from GitHub into eve's native skill path:

npx skills add Ambush-AI/ambush-stream-skills \
  --skill manage-ambush-streams \
  --agent eve \
  --yes

Then create a Vercel Connect connector from the linked eve project:

npx vercel@latest connect create https://api.ambush.ai/mcp \
  --name ambush-streams \
  --yes

That service and name produce the connector UID api.ambush.ai/ambush-streams. Use it in agent/connections/ambush-streams.ts:

import { connect } from "@vercel/connect/eve";
import { defineMcpClientConnection } from "eve/connections";

export default defineMcpClientConnection({
  url: "https://api.ambush.ai/mcp",
  description:
    "Ambush Streams: create and manage personalized real-time news streams, review emissions, and route future events to connected destinations.",
  auth: connect("api.ambush.ai/ambush-streams"),
  approval: ({ toolName }) =>
    toolName === "delete_feed" ? "user-approval" : "not-applicable",
});

The generic Ambush OAuth connector currently supports user-scoped subjects, so each eve user authorizes their own Ambush account. Do not switch this connection to an app-scoped principal: headless runs without an authenticated user are not yet supported by this connector.

Example requests

• "Create a stream for material cybersecurity incidents affecting Canadian banks."
• "Pause my AI regulation stream."
• "Show the five latest items from my semiconductor supply-chain stream."
• "For every event from that stream, produce a trade thesis or no-trade result and send it to my Trade Ideas Slack channel."

Included skills

• manage-ambush-streams: create, inspect, update, process, route, pause, resume, and permanently delete streams, and review emitted news items.