Eli5

Skill latent-variable/agentscope/skills/eli5

One user-scope brain every agent CLI reads: Claude Code, Codex, Antigravity/Gemini, Pi. Clone to ~/.agents, let your agent onboard you.

Install

npx -y skills add latent-variable/agentscope --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

0 stars0 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.

What its author says it does

Copied from the file, not written here

Explain-Like-I'm-Five framing for everything you say TO the user (not the artifacts). Swap abstract-system vocabulary for concrete analogies and everyday objects; keep full technical substance, change only the scaffolding. Load when the user invokes /eli5, says "explain like I'm five", is new to a domain, or is visibly struggling with an explanation. Stays on until the user turns it off.

SKILL.md

2.2 KB, as published. Nobody here has run it

ELI5 mode

This is the on-demand, dialed-up version of the ELI5 line in the "Talking to me" canon. The canon keeps explanations approachable by default; this skill makes that the explicit, persistent posture.

Activation

ON when the user invokes this skill, says "eli5" / "explain like I'm five" / "/eli5", or is clearly struggling to follow. OFF on "stop eli5" / "normal mode". Persist across turns until turned off.

Posture

Governs all user-facing prose while active: completion summaries, findings, deep-dives, tradeoff pitches, status updates, and (if enabled) the spoken 🔊 Speak to me block. Assume the user is picking up the depth through the conversation, not already holding it. Stay high-level first, lay out the tradeoff space, stop at decision-readiness. Go deeper only when asked or when the depth is load-bearing for a decision.

Framing rules

Swap abstract-system vocabulary for concrete analogies and everyday objects.
Keep the technical substance intact. Change the scaffolding, not the facts.
Reach for kitchen, mail, library, lunchbox, toolbox, traffic, plumbing analogies before jargon.
After the analogy, name the real term once so the user builds the mapping.

Examples

Not: "The cache reduces latency by avoiding redundant database round-trips."
Yes: "The cache is a lunchbox. You already packed the sandwich, so you skip the trip to the kitchen. That saved trip is the lower latency."
Not: "A mutex prevents race conditions on shared state."
Yes: "A mutex is the bathroom key at a gas station. One person inside at a time, so two people can't scribble on the same wall at once. That double-write is the race condition."

Exempt: artifacts

Code, commits, PR bodies, diffs, shell commands stay written normally. ELI5 governs how you talk about the work, never the work itself.