Skip to content
Agentmart
Submit a tool

daily-brief

Collect dated evidence, write and check one morning note, rehearse without live side effects, and explain the exact delivered version.

Skill by MichaelZelbel

IndexedMIT
Write a review

Install

Works with Claude Code, Claude apps, Codex, Cursor, Gemini CLI, GitHub Copilot, VS Code, OpenCode, Goose, Windsurf. Read the repository before you give any tool access to your data.

npx agentmarket install skill/michaelzelbel-daily-brief

Asks which of your tools to add it to (Claude Code, Codex, Cursor, VS Code, Windsurf, Gemini CLI, Claude Desktop) and shows the change first. See https://agentmart.dev/cli.

Get the weekly diff

The weekly diff: one email a week with new agent tools, notable updates and grade changes, plus security alerts for tools you follow. Double opt-in, one-click unsubscribe.

We use your email only for this newsletter. Privacy

About

Collect dated evidence, write and check one morning note, rehearse without live side effects, and explain the exact delivered version.

Use cases: writing, productivity

No reviews yet. Installed it? Reviews come from verified installers only. How reviews work

Permissions

What this skill can do and what it asks for, from its SKILL.md. Checked Oct 25, 2026.

Tools
No tools found
Secrets
None asked for
Install scripts
None found

Network hosts it talks to 1

  • codeload.github.com · found in install command

Files it touches

  • ~/.claude/skills/daily-brief

What's inside

Skill name
daily-brief
Description
Collect dated evidence, write and check one morning note, rehearse without live side effects, and explain the exact delivered version.
License
MIT
Instructions
45 lines in SKILL.md · view on GitHub

Files (1)

  • SKILL.md

SKILL.md (first lines)

# Daily brief

Use one maintained path for collection, writing, checking and delivery. Multiple writers and a fallback sender must never independently send the same morning. The morning-note recipe owns the words; mc-brief-operations owns decisions, links and explanations.

## Selected installation and entry points

Read its operating manual, voice, current goals and delivery authorization. Confirm the schedule owner and timezone through normal Settings. Use the normal Routine selector for morning-brief or brief-rehearsal and the normal due-routine control. A rehearsal must send nothing and must operate on a separate retained copy, including reminder completion and source bookkeeping.

Keep results in the notebook unless the user has explicitly enabled the exact delivery connector. A configured credential is not permission to send. Enabling or running a routine must preserve unrelated deliberately paused jobs, other installations and company systems.

The product implementation is notebook/core/briefing.mjs with shared scheduler receipts. Do not invoke private personal-engine executables. A missing stage is a missing implementation; do not disguise it as a missing optional account.

## One path, four stages

1. **Collect before writing.** Read the adopted goals and attention plan, today's saved decision, open obligations, recent actual work, the user's recent words, health observations, configured calendar receipts, watch findings and primary-source articles. Refresh the selected permitted sources. Retain dates, full relevant source text, successful read evidence and source gaps. A failing calendar/news source costs its part, not the whole morning. A title or search snippet is not an opened article.
2. **Write facts, then one note.** Apply the common card cap before passing items to the writer. Read user replies back into the cards first. Include current health counts computed from the selected measurements, verified completed work, prior shown/answered/rejected cards, earlier news and the latest rehearsal findings. The morning-note method produces one note and a used-card list. A separate maintained lead workflow prepares any social item; the morning writer must not improvise it.
3. **Check the exact text.** Mechanical gates check links, dates, numbers, rule/voice restrictions, IDs, repeats and exact numbered choices. An independent plain-words reader checks whether a person seeing none of the work can act. A fact check compares every story with its opened source. The writer gets one bounded rewrite with the actual failure reasons. Any smallest mechanical repair must retain meaning. Independent final judges may cut unsupported passages; they may not invent replacement facts. Insert an already verified social card by code, then rerun the mechanical gate on the final exact payload. An invalid result stays undelivered with retained evidence.
4. **Save, then deliver once.** Save the exact final note, source/check records, used cards and delivery parts before sending. Keep draft preparation, a checked notebook result, an attempted send and an acknowledged delivery distinct. Mark only the actually delivered items shown. A page or auxiliary link failure must not turn an otherwise valid note into a fictional delivery success.

## Preview and rehearsal

A preview reads the same evidence and uses the same checks but sends nothing. A rehearsal runs those stages on a copy with the software intended for the next morning. It must not close a real obligation, mark a real watch finding shown, advance a real news cutoff or create live delivery notifications.

Keep the copy and verdict. Record whether native file sources, selected connectors and dependencies were available. A rehearsal that silently lacks a required source is incomplete. File one quiet repair for a crash, no valid note or unreasonable runtime, then feed the actual finding to the morning writer. Repeated failures should share one repair identity and bounded attempts.

## Delivery safety

One scheduler owner controls delivery. Store a stable delivery identity and exact payload before the connector is called. Accept success only from the selected connector's actual acknowledgement. If a send may already have happened, stop automatic replay and retain the uncertainty for review. Resume only parts known not to have been attempted. An explicit resend is a separate authorization.

The keeper owns recovery after a failure: use the newest verified saved parts or an honest actionable notice. Never send a stale preview as though it were the original morning note. Do not make the user approve an ordinary reversible repair the assistant is already authorized to perform.

## Explaining a note later

Locate the actually delivered version from its delivery receipt, then read its saved facts, source evidence, judge results and exact text. A later preview is not what the user received. Answer from the original evidence; re-verify a current claim when needed. Preserve the delivered version and any correction as separate history.

## Acceptance

Exercise a thin news pool, missing calendar, new health evidence, a numbered approval list, a failed check repaired once, an unrepaired failure, an unchanged repeat, a delayed acknowledgement and restart. Check ordinary notebook controls and actual retained state. A source test or a saved note alone does not establish readable receipt, reliable scheduling or completed delivery.


## Installed personal workspace

Use the current user workspace and its configured providers. Keep original workflow, command contracts, scripts and verification criteria. Read the workspace authorization rules before sends, sign-ins, payments or publishing. Search existing device-private credentials before asking for configuration. Saved output and a passing screen are not evidence that the full requested result happened.

From MichaelZelbel/godspeed-mission-control, MIT.

More from MichaelZelbel/godspeed-mission-control

Signals

Stars
0
Forks
0
Open issues
10
Last commit (this folder)
Oct 18, 2026 (10 days ago)
Last release
Oct 24, 2026 (4 days ago)

From GitHub (fetched Oct 24, 2026).

Details

License
MIT
First seen
Oct 24, 2026
Last checked
Oct 24, 2026

More from MichaelZelbel

Cite this page

daily-brief skill. Agent Mart. https://agentmart.dev/skills/michaelzelbel-daily-brief (last updated 2026-10-24)