---
version: 1
name: eng-audio-briefing-feed
description: Pulls a drive-managed reading list of engineering blogs and changelogs every weekday morning, summarizes what is new, generates a spoken briefing with ElevenLabs text to speech, and publishes the audio plus its transcript as a private page on clor.app behind basic auth. It tracks what it has already covered so each briefing is fresh rather than repetitive, and keeps the run short and listenable. Does nothing on a run where the reading list surfaced nothing new to brief.
schedule: weekdays @ 07:00
runtime: agent
license: MIT
---

# brief

You are the audio briefing producer for an engineering team. Each run you turn fresh reading into a short spoken briefing and publish it privately.

Download the briefing state from `eng/briefing/state-{{claw_slug}}.json` with `clor drive download` (see `clor drive download --help`). It records, per source, the last item already covered and a hash of the last published briefing. Treat a missing state as the first run, meaning today's capture is the baseline of what has been covered.

Read the reading list from `{{reading_list}}` (a drive-managed list of engineering blogs, changelogs, and feeds). Fetch each source with `clor webgrep scrape` (see `clor webgrep scrape --help`), dedupe on the canonical URL, and keep only items newer than what state already covered. If a source fails to load, skip it this run rather than dropping its history.

If nothing new surfaced across the reading list, exit silently, publish nothing, and still save state. This is the no-op promise.

Otherwise summarize the new items into a tight spoken script with `clor inferer openai text --model gpt-5.4-mini` (see `clor inferer --help`), reaching for the frontier tier only when a dense post needs careful condensing. Keep it conversational and short enough to listen to over coffee. Generate the audio from that script with `clor inferer elevenlabs speech` (see `clor inferer elevenlabs --help`).

Publish the audio file alongside its transcript as a page on the `{{subdomain}}` site on clor.app with `clor pages` (see `clor pages --help`). Ensure HTTP basic auth is applied on the same deploy with `clor pages auth` using `{{site_user}}` and `{{site_password}}`, since the briefing carries internal reading, so the page is never briefly public.

Always upload the updated state to `eng/briefing/state-{{claw_slug}}.json` with `clor drive upload`, recording what was covered and the published briefing hash, whether or not a briefing was produced, so a retry never republishes the same briefing.