Auto-draft your weekly changelog post from real data

Generate a 5-tweet thread (or Slack post, or LinkedIn update) summarizing the week's product metrics. Real numbers, your voice, 5 minutes.

Published

difficulty beginner · time to value 3 minutes · tools used 3

Just say this

Generate a Twitter/X thread summarizing this week's metrics: signup count, retention delta vs last week, top error we fixed, biggest user. Make it sound real, not corporate.

Why this matters

Weekly content is the #1 organic growth lever for indie SaaS, and the #1 thing founders skip. The reason is always the same: “I don’t know what to post.”

You actually do know — you just need someone to translate your week’s numbers into a story. Old way: scroll three dashboards, screenshot a chart, write a tweet, edit it 6 times, post. 90 minutes for one post.

New way: prompt the agent with the numbers you care about and the voice you want. Get back a draft. Edit two lines. Post. 5 minutes flat. Compounds over months into thousands of organic visitors.

What you get

  • A 4–6 tweet thread (or Slack post / changelog entry / LinkedIn update — your choice)
  • Real numbers pulled from your data
  • A voice you can iterate (the agent learns your tone the more you use it)
  • Optionally: a chart screenshot URL from a published query

Walk through it

You

Generate a Twitter/X thread summarizing this week’s metrics: signup count, retention delta vs last week, top error we fixed, biggest user we got. Voice: matter-of-fact, slightly self-deprecating, no emojis except where genuinely warranted. End with a soft CTA to try the product.

Agent

Let me pull the numbers first.

agentry_run_recipe {
  recipe_id: "weekly_metrics",
  params: { project_id: "default", week_ending: "2026-05-16" }
}
{
  "signups": { "this_week": 142, "last_week": 118, "delta_pct": 0.203 },
  "activation_rate": { "this_week": 0.41, "last_week": 0.38, "delta_pct": 0.079 },
  "day_7_retention": { "this_week": 0.62, "last_week": 0.58, "delta_pct": 0.069 },
  "top_resolved_case": {
    "id": "c_f72a",
    "message": "TypeError on checkout for free-tier users",
    "users_affected_before_fix": 14
  },
  "biggest_new_signup": {
    "email": "[email protected]",
    "signed_up_at": "2026-05-14",
    "company_domain": "anthropic.com"
  }
}
Agent

Draft thread:

1/  Week in numbers.

    142 signups (up 20%), day-7 retention 62% (up from 58%),
    activation rate 41% (up from 38%).

    The retention bump is from the onboarding email we shipped Monday.
    Going to keep ripping that thread.

2/  Bug of the week: 14 free-tier users were silently hitting a
    TypeError on checkout. Their fault for clicking the upgrade
    button, I guess.

    Fix shipped Tuesday morning. Apologies to everyone who hit it
    and didn't tell us.

3/  Biggest new signup: someone from anthropic.com signed up on
    Wednesday. We assume it's Claude lurking.

    If you're reading this, hi.

4/  Tomorrow's plan: ship the team-invite flow. Today's plan:
    finish writing it.

5/  agentry.sh — error monitoring + analytics for teams that
    would rather ask an agent than open another dashboard.

Two things I held back from: I almost wrote “we’re so back” in tweet 1 but you said no corporate-speak — that includes meme-speak. Also the bug fix line is borderline self-deprecating; let me know if it should be straighter.

The output

The agent returns a paste-ready thread. You edit voice (5%), add a screenshot if you want, post. Total time: 5 minutes.

If you want screenshots too, ask the agent to publish the relevant chart:

Add a chart of the retention curve as tweet 2’s media.

It publishes the query, gives you the URL, and you screenshot it (or embed an iframe).

Setting it up

Two ingredients:

1. Consistent event names for the metrics that matter to you. The weekly_metrics recipe expects events like signup_completed, first_action_completed (for activation), and any retention proxy event. Names don’t have to match exactly — the agent finds them via agentry_list_event_names. But they have to exist.

2. (Optional) A voice_note saved in memory. After the agent drafts a few posts, save the rules:

agentry_remember "Voice rules for changelog posts:
  - No emojis except when genuinely fitting
  - Slightly self-deprecating, never mean
  - Numbers are the spine — one chart per post if applicable
  - End with a soft CTA, never aggressive"

Next time the agent drafts a post, it pulls these rules automatically.

Variations

  • “Generate a LinkedIn post version of the same content — more professional, less terse.”
  • “Draft a Slack message version for #company-all, focusing on what changed for customers.”
  • “Generate a monthly investor update from the last 4 weekly summaries.”
  • “Set up a Routine: every Friday at 5pm, draft the weekly thread and DM it to me in Slack.”

Try this recipe in your own agent.

Paste the prompt above into your agent. It'll set up Agentry against your data, then run the recipe and show you the real output.

Install https://agentry.sh/agentry.md and set it up