Eli5

Explain current work plainly for a teammate without recent context, defining mechanisms and removing task-specific shorthand.From its SKILL.md

Install

npx -y skills add beorn/bearly --skill eli5

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

3 stars3 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

SKILL.md

2.7 KB, as published. Nobody here has run it

/eli5 — Explain Like a Teammate Just Walked In

Read /agent-system before editing this doc.

The user has only general project knowledge — what the project and its top-level packages ARE. Nothing else from the last N minutes of your work. Reset to that floor.

The 4 rules

Mechanism, not option name. ❌ "we could enable singlePassLayout." ✅ "the framework currently runs layout → re-renders → runs layout again until stable. The option would force one pass."
Define names on first use. "the cascade", "the feedback loop", "Suspect 3", Content.tsx — gloss on first mention even if it's the 20th time you've thought it.
Numbers, not adjectives. "200 violations", not "still many". "1 second", not "briefly". "70% drop", not "big improvement".
One screen, ~25 lines. If it doesn't fit, drop secondary detail. Pick the load-bearing parts.

Structure (when explaining problem + options)

What's happening:   mechanism — 2-3 sentences
What I did:         change + concrete result
What's left:        remaining mechanism
Options, plainly:   (a) <name> — what it changes, tradeoff
                    (b) ...
Recommendation:     1 sentence

For a simple question, just write prose under "What's happening" and stop.

Before / after (a real one)

❌ "Should I (a) try singlePassLayout, (b) hunt the 336-wide text, or (c) audit useBoxRectDangerously call sites? Suspect 3 didn't pan out and AsideLayout was reverted."

✅ "The remaining churn is the rendering framework doing layout, then re-rendering, then doing layout again until things settle — call this the cascade. Three angles:

Force one layout pass — the framework has a setting that runs layout exactly once. Risk: components that read their own size get 0 on first paint.
Find the long unwrappable text — one chat element measures 336 columns wide inside an 85-column parent. Adding wrap="wrap" fixes that one leaf.
Reduce measurement reads — every measurement-hook call is a re-render trigger. We have 3; some may be redundant."

Self-trigger when you catch yourself

Writing any of these in user-facing text → stop, run /eli5:

"the X fix" / "the Y change" — name what changed mechanically
"Suspect N" / "Phase N" / "Attempt N" — your enumeration isn't theirs
"Per the diagnosis" / "Per the bead" — they haven't read it
A bare file path — gloss it ("Content.tsx — the chat-layout file")