Skip to content
Back to skills

Daily Brief

ASecurity

Compose the operator's daily brief — schedule, things waiting on them, ongoing work, and heads-ups. LOAD THIS SKILL FIRST (load_skill) and deliver the brief as an HTML artifact via show_artifact — never as chat text. Run only when the operator explicitly asks for their brief (or /daily-brief), or when executing a scheduled brief task. An ordinary question about today's calendar is answered directly, not with a brief.

  • 11 stars
  • 0 votes
  • 0 copies
  • 2 views
  • Added October 2, 2026
ai-agentsgo

Security analysis

A100/100

Scanned October 2, 2026

npx -y skills add protoLabsAI/protoAgent --skill daily-brief --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Daily Brief?

Add the live security badge to your README. It updates with every re-scan.

Security grade badge for Daily Brief
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/protolabsai-daily-brief/badge)](https://www.skillsdirectory.com/skills/protolabsai-daily-brief)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

Download with Pro
SKILL.md
---
name: daily-brief
description: Compose the operator's daily brief — schedule, things waiting on them, ongoing work, and heads-ups. LOAD THIS SKILL FIRST (load_skill) and deliver the brief as an HTML artifact via show_artifact — never as chat text. Run only when the operator explicitly asks for their brief (or /daily-brief), or when executing a scheduled brief task. An ordinary question about today's calendar is answered directly, not with a brief.
user_facing: true
slash: daily-brief
---

# Daily brief

One page that answers three questions before the first coffee is finished:
*what's my day shaped like, who is blocked on me, and what was I in the middle
of?* Tell the operator it takes a couple of minutes, then work quietly and
deliver the page.

## Take stock of sources

Use what's connected; skip what isn't — the brief adapts rather than
apologizing. In order of value:

1. **Memory** — `memory_recall` for the operator's active projects, stated
   priorities, and preferences. This is what makes the brief *theirs*: an
   event titled "sync" means nothing until memory says which project it
   belongs to.
2. **Calendar** — today's events; glance at tomorrow morning only to decide
   whether today needs prep time for it.
3. **Email** — conversations where the operator owes a reply: someone asked
   *them* specifically and the thread has gone quiet on their side. A thread
   where anyone on a list could answer doesn't count. If nothing qualifies,
   fall back to what's unread and recent, clearly labeled as such.
4. **Overnight work** — anything that ran unattended since yesterday:
   `list_schedules` for what fired, and scheduled/background results are
   indexed into memory, so `memory_recall` surfaces what a run produced.
   Something finished while the operator slept belongs on the page; a run
   that failed belongs higher.
5. **Work folders** — files modified in the last day or two in the fenced
   folders: yesterday's deliverables are today's follow-ups.
6. **Notes / tasks** — anything explicitly marked for today.

Budget the calls: one pass per source, no fishing expeditions. A brief that
takes ten minutes to gather has already failed its job.

## Decide what makes the page

Everything competes for a small page. Keep an item only if the operator would
act on it *today*: a meeting they must prepare for, a person waiting on them,
work mid-flight with a deadline in sight. Fold the rest into one quiet line or
drop it. Never pad — a light day should *look* like a light day; that is
information too.

Order: the day's shape first (schedule with prep flags), then **waiting on
you** (people first, oldest ask first), then **in flight** (ongoing work with
its next action), then a short **heads-up** (tomorrow's early meeting, an
approaching deadline, a stale promise memory knows about).

## Render

The brief IS the artifact — always deliver it with `show_artifact(kind="html",
...)`, never as chat markdown, even on a light day (a light day gets a light
page; the chat reply is one pointer line, "Your brief is in the panel").
Self-contained: system font stack, no external assets or network fetches,
readable in light and dark.
Design it like a well-set newspaper front page, not a dashboard — a dated
masthead, one strong opening line summarizing the day in plain words, clear
sections, generous whitespace. Timestamps in the operator's timezone; every
item phrased as a decision or action, never raw data dumps.

If a writing-voice profile exists (`my-writing-style`), the prose follows it.

## Recurring setup

If the operator asks for this every morning (or any cadence): use the schedule
skill's discipline — distill into a self-contained prompt that names the
sources to check, the fenced folders by path, the delivery form (HTML artifact
via show_artifact), and the language to write in (the language the operator
used when asking). protoAgent cron is unrestricted: "weekdays at 7" is
`0 7 * * 1-5`, but Sunday-evening week-ahead briefs or first-of-the-month
reviews work just as well.

## Ground rules

- Explicit invocation only — never volunteer a brief mid-conversation.
- Read-only gathering: composing the brief must not send, reply, archive, or
  modify anything.
- If every source is disconnected, say so and point at Settings ▸ Plugins —
  don't render an empty page.

Attribution

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments

Loading comments…

Newsletter

Get the weekly list of safe new skills, and the ones we pulled.