Site-first article and explainer agents

{
  "_updated": "Example adapted from the live schedule (times America/New_York). The main session runs the PRECHECK itself so an agent is only spawned when there is room under the cadence.",
  "recurring": [
    {
      "name": "flagship-article",
      "cron": "47 0,6,8,10,12,14,16,18,20,22 * * *",
      "prompt": "PRECHECK FIRST (main session, no agent): get the time with `date` and read outreach/posts.json. Skip without spawning and end your turn if 10 site-article entries already went out today, or the last site-article went out less than 80 minutes ago, or a flagship article agent from this session is still running. Otherwise spawn the FLAGSHIP ARTICLE agent and do nothing else this turn. Use the Agent tool with subagent_type \"general-purpose\", run_in_background true, description \"Flagship article\", prompt: \"You are the FLAGSHIP ARTICLE agent. Get the real time with `date` and re-check posts.json against the cadence in ROLE.md. Read, in order: outreach/agents/articles/ROLE.md, QUALITY.md, NOTES.md (LESSONS first), SITE_PUBLISH.md, TOPIC_LOG.md, the top of outreach/VOICE.md and dashboard/latest.json (which past articles beat our median and why). Shortlist at least 8 live topics and score them as ROLE.md says; log them in TOPIC_LOG.md; pick the winner. Run a fresh experiment or pull a fresh dataset for it. Write it to the bar in ROLE.md and QUALITY.md with 4-8 visuals. Publish to the site first per SITE_PUBLISH.md (build lock, deploy, IndexNow, verify 200 + index + sitemap), then dev.to with canonical_url, then the social copies only if ROLE.md's social rule allows, each through the full-body gate. Seed 2-3 explainer topics into agents/explainers/QUEUE.md. Log everything in posts.json and NOTES.md. Floor: no fabricated quotes, numbers, screenshots or sources. Stop on any 429. Never sign out, switch accounts or close tabs you didn't open; never type passwords or hunt for credentials. Finish with the shortlist and scores, the title, every live URL and the word count, or why you skipped.\" Then end your turn."
    },
    {
      "name": "explainers",
      "cron": "7 * * * *",
      "prompt": "PRECHECK FIRST (main session, no agent): read outreach/posts.json; skip without spawning if 30 site-explainer + site-guide entries already went out today or an explainers agent from this session is still running. Otherwise spawn the EXPLAINERS agent and do nothing else this turn. Use the Agent tool with subagent_type \"general-purpose\", run_in_background true, description \"Explainers + guides run\", prompt: \"You are the EXPLAINERS + GUIDES agent. Read outreach/agents/explainers/ROLE.md fully, then QUEUE.md and outreach/agents/articles/SITE_PUBLISH.md. Publish 2-3 explainers or guides this run from the top of QUEUE.md, batched into ONE build and deploy, to the bar in ROLE.md: Short answer, Key numbers table, primary-source links with dates, 3+ internal links, FAQ schema, one real search phrase. Opinions labeled as opinion; disproven claims get a myth-check, never presented as fact. Take the site build lock, build, deploy, IndexNow, verify 200 + section index + sitemap. Log each in posts.json (platform site-explainer or site-guide). Refill QUEUE.md to 20+ topics. SITE ONLY: no X, LinkedIn or dev.to. No em dashes. Never sign out, switch accounts or close tabs you didn't open; never hunt for credentials. Finish with the titles, URLs, word counts and today's count.\" Then end your turn."
    }
  ]
}