emoji-readout
Skill8 starsChanged 5 months agoMITView on GitHub
Write a short status readout of long-running work as a handful of emoji-led bullets a person can scan in five seconds after walking away: the time and how long it has been, what moved, what is stuck, and, last, whether anything needs them (β) or not (π’). Use when asked for "a status update", "an emoji summary", "a readout", "where are we", "updates every N minutes", or at the checkpoints of a burndown, a long build, a migration, or any unattended run someone checks in on. Ships a small renderer (`lib/readout.ts`) for drivers that print the same shape from code. Skip for a one-line answer, and for a final report whose reader needs prose and reasoning rather than a scan.
In this repository23 other files it gives its agents.
- skills/asset-optimization/AGENTS.mdAGENTS.md
- android-emulator-harnessSkill
- android-emulator-mask-testingSkill
- appraise-github-issuesSkill
- asset-optimizationSkill
- burn-down-github-issuesSkill
- carve-github-issueSkill
- code-github-issueSkill
- expo-developer-toolsSkill
- expo-ios-simulatorSkill
- fix-github-issueSkill
- how-to-planSkill
- ios-simulator-flow-evidenceSkill
- ios-simulatorSkill
- ios-simulator-triageSkill
- mobile-accessibilitySkill
- playwright-camera-mask-testingSkill
- playwright-gif-captureSkill
- playwright-harnessSkill
- playwright-storybook-flowsSkill
- prove-github-issueSkill
- prove-work-on-githubSkill
- zone-composerSkill
The file
We split this skill into the rules it gives an agent
1 rule. About 0 are common, 1 is distinctive, and 0 are very specific to this project.
β0
Common Β· estimate
Many files like this have a rule of this kind.
β1
Distinctive Β· estimate
This author's own choices. Worth a look.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 51
Calls made without asking: Every consequential choice the agent made on its own (a skipped test, a chosen default, a deleted branch), and β¦
β βCommonβ is an estimate: at least 1 in 5 of the 372,429 skill files we read have a rule of the same kind, not the same words. βVery specificβ is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Emoji Readout
Someone walked away and came back. In five seconds the readout tells them whether they are needed, and if not, what changed and whether it is still moving. It is a scan, not a story: high level, list-like, one line per thing, an emoji leading each line so the eye can sort it before reading it.
16:56 CDT Β· 1h 12m in
- β
PR #101 merged; checks green (5 pass, 6 skip)
- π def5678 in tests, 12 of ~18 min
- π§± #214 stuck 25 min: the worker keeps failing the lint gate on the same file
- π©Ί all processes alive; dev server answers
- π
next readout ~17:11
- β Needs you: may I delete the superseded plan file? (work continues meanwhile)Always there
- The time, and how long. The clock time with its zone, read from the clock (
date) every time, never estimated; then how long the work has been running (1h 12m in). When the gap since the last readout matters too, say it in its own words (20m since last), never asin. - What moved. What changed state since the last readout (landed, started, finished, failed,
closed), and anything still running normally, with its progress against what is normal for it
(
12 of ~18 min). If nothing moved, say so:nothing moved since 16:40. A quiet readout must never look like one nobody checked. - What is stuck. Anything not moving that should be: stalled, retrying, failing the same way, waiting on something outside, with how long. A stuck thing repeats in every readout until it moves; it is the exception to reporting only changes.
- Whether you are needed, last. β means one thing only: the reader has to decide, permit, or
do something. Each item is answerable in one reply (a yes or no, lettered options, or a short
question), and it repeats until answered. When nothing needs them, the last line is
π’ Nothing needs you., so the emoji alone says which, and silence is never ambiguous. Never use β for anything else, and never bury a question elsewhere.
When it applies
- Calls made without asking. Every consequential choice the agent made on its own (a skipped test, a chosen default, a deleted branch), and whether it can still be undone. Never optional when there is one.
- The next checkpoint. For a run on a cadence or left unattended: when the next readout is due
(
next ~17:11), and separately, if known, an estimate of when the work finishes. A missed checkpoint is how a reader learns the run died. - Run state, when it is not simply running. Paused, finished, or stopped early, and whether work continues while a β waits.
- Progress toward done. For work with a known size, the count and what it counts:
7 of 12 issues merged,3 of 5 verified on staging.
The agent's call
Everything else is judgement about what this reader needs from this work right now.
- Report what matters; omit what is inconsequential. A section with nothing worth saying is absent, not "none". A risk, a slow step, a surprise, a number that moved: include it if it would change what the reader thinks or does.
- Pick emoji that read at a glance, one per line or group. Some that tend to work: β done, π moving, π§± stuck, π©Ί health, β οΈ worth watching, π§ͺ tests, π deploys, π§Ή cleanup. Invent one when the work has its own category. Within one session keep an emoji's meaning stable, so two readouts compare.
- Group only when it helps. A few peers under one emoji line read faster than a paragraph; most readouts need no nesting at all.
What keeps it trustworthy
- Receipts, briefly. A short SHA, a PR or issue number, a count, a check name, a log time. "Tests pass" is a claim; "412 of 412 at 16:51" is a receipt.
- Re-observe; don't recite. Build each readout from fresh checks, not memory of the last one. An older fact keeps its time ("green at 16:40").
- Say what is inferred. Observed facts plainly; anything inferred or guessed says so.
- Short. Around ten lines is plenty, fifteen a ceiling. No preamble, no closing offer of next
steps. If nothing moved, nothing is stuck, and nothing needs the reader, the readout is the
header,
nothing moved since HH:MM, andπ’ Nothing needs you.
On a cadence ("every 15 minutes"), schedule it with the harness's recurring trigger, build each one fresh, and say in the last one that it is the last.
From code
A driver that prints readouts imports the renderer and supplies its own sections:
import { renderReadout } from '<this-skill-dir>/lib/readout.ts';
console.log(renderReadout({
since: runStartedAt, // header shows "Β· 1h 12m in"
title: 'floor walk',
sections: [
{ emoji: 'β
', lines: ['PR #812 present on the deployed site (look)'] },
{ emoji: 'π©Ί', label: 'Health', lines: ['up: / 200 in 120ms; deployed a1b2c3d'] },
],
needsYou: [],
}));Sections print in the order given; blank lines are dropped, and so is a section left with none.
A section with one line prints on one line, labelled or not. Several lines nest under the label;
without a label, each prints as its own bullet behind the emoji. A section given β is moved into
needsYou, which always prints last, or as π’ Nothing needs you. when empty.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

