Cron prompt hardcodes Twitter usernames, ignoring dynamic sources DB

[pathu9796-cpu]

Bug Description

When using ClawFeed as an OpenClaw skill with a cron job, the agent prompt hardcodes a fixed list of Twitter usernames to fetch. This means any sources added/removed via the database are ignored — the cron agent never reads the actual sources table.

Root Cause

The default cron prompt tells the agent to fetch from a static list like @karpathy @sama @nvidia..., rather than first calling generate-digest.mjs to get the dynamic source list from the DB.

Steps to Reproduce

1. Add a new Twitter source via the DB (e.g. @insane_analyst)
2. Run the cron job
3. The new source is never fetched — the agent uses its hardcoded list instead

Expected Behavior

The cron/agent prompt should:

1. First run node scripts/generate-digest.mjs to get the live source list
2. Iterate over all sources dynamically by type
3. For twitter_feed sources, use: agent-reach search-twitter "from:{username}" -n 10

Suggested Fix

Update the default cron prompt template to use the dynamic source list. The correct agent-reach command for Twitter timelines is:

agent-reach search-twitter "from:insane_analyst" -n 10

Note: agent-reach read and agent-reach timeline do NOT work for user timelines — only the search-twitter from: syntax is functional.

Environment

• ClawFeed version: 0.8.1
• Runtime: OpenClaw skill (cron job mode)