Skill Authoring

Skill Authoring

A skill exists to wrangle determinism out of a stochastic system. Predictability — the agent taking the same process every run, not producing the same output — is the root virtue; every lever below serves it.

Bold terms are defined in GLOSSARY.md; look them up there for the full meaning.

Credits: Based on writing-great-skills by Matt Pocock (https://github.com/mattpocock/skills). Adapted for the Lemon AI Hub plugin ecosystem.

Invocation

Two choices, trading different costs:

• A model-invoked skill keeps a description, so the agent can fire it autonomously and other skills can reach it (you can still type its name too). It contributes to context load — the description sits in the window every turn. Mechanics: omit disable-model-invocation, and write a model-facing description with rich trigger phrasing ("Use when the user wants…, mentions…").
• A user-invoked skill strips the description from the agent's reach: only you, typing its name, can invoke it — and no other skill can. Zero context load, but it spends cognitive load: you are the index that must remember it exists. Mechanics: set disable-model-invocation: true; the description becomes human-facing — a one-line summary, trigger lists stripped.

Pick model-invocation only when the agent must reach the skill on its own, or another skill must. If it only ever fires by hand, make it user-invoked and pay no context load.

When user-invoked skills multiply past what you can remember, that piled-up cognitive load is cured by a router skill: one user-invoked skill that names the others and when to reach for each.

Writing the description

A model-invoked description does two jobs — state what the skill is, and list the branches that should trigger it. Every word increases context load, so a description earns even harder pruning than the body:

• Front-load the skill's leading word — the description is where it does its invocation work.
• One trigger per branch. Synonyms that rename a single branch are duplication — "build features using TDD … asks for test-first development" is one branch written twice. Collapse them; keep only genuinely distinct branches.
• Cut identity that's already in the body. Keep the description to triggers, plus any "when another skill needs…" reach clause.

Information hierarchy

A skill is built from two content types — steps and reference — that mix freely: a skill can be all steps, all reference, or both. The core decision is which to use and where each sits on the information hierarchy, a ladder ranked by how immediately the agent needs the material:

1. In-skill step — an ordered action in SKILL.md, the primary tier: what the agent does, in order. Each step ends on a completion criterion, the condition that tells the agent the work is done. Make it checkable (can the agent tell done from not-done?) and, where it matters, exhaustive ("every modified model accounted for").
2. In-skill reference — material the agent consults as needed, secondary to steps. Lives in the same file but is not a sequence; the agent dips in when a step demands it.
3. Disclosed reference — pulled out of SKILL.md behind a context pointer (a link or named callout). The agent only loads it when a step or branch needs it, keeping the top tier lean.

Move material down the ladder (from in-skill step, to in-skill reference, to disclosed) until each tier carries only what its rank demands. The test for "is this a step or reference?" is sequence: if order matters, it's a step; if the agent consults it on demand, it's reference.

Completion criteria drive legwork

A demanding completion criterion drives thorough legwork — the digging the agent does within the work — whether the skill has steps or not, since "every rule applied" binds flat reference just as "every step done" binds a sequence.

Leading words

A leading word is the compact term that anchors a skill's invocation and its execution — the single word you would actually type. It does triple duty: it names the skill (so you can reach it), seeds the description's trigger phrasing, and concentrates what the skill is in the fewest tokens.

• A weak leading word (be thorough when the agent is already thorough-ish) is a no-op — it changes no behaviour.
• A strong leading word (relentless) shifts behaviour without changing technique.
• The leading word belongs at the front of the description: that is where invocation happens.

Granularity

Granularity is how finely you divide skills, and each cut spends one of the two loads, so split only when the cut earns it. Two cuts:

• By branch — split when one branch of a skill grows distinct enough to deserve its own leading word and description. Lowers the context load of the parent (the branch's triggers move out of the parent's description) but adds cognitive load if the split becomes user-invoked.
• By sequence — split an ordered skill when later steps only matter after earlier ones finish, and the agent rushes to completion. Hiding post-completion steps behind a real context boundary (a user-invoked hand-off or a subagent dispatch) cures premature completion; an inline model-invoked call leaves the later steps in context and clears nothing.

Steering

Steering is how a skill shapes the agent's runtime behaviour beyond the sequence of steps. Two disciplines:

• Keep each meaning in a single source of truth: one authoritative place, so changing the behaviour is a one-place edit.
• Prefer positive steering over negation. Don't think of an elephant names the elephant and makes it more available, not less. State the target behaviour so the banned one is never spoken; keep a prohibition only as a hard guardrail you can't phrase positively, and even then pair it with what to do instead.

Pruning

Pruning is the discipline that keeps a skill lean. Every line costs tokens and maintenance, so each must earn its place. The four pressures to cut against are the failure modes below.

Keep the single source of truth: one authoritative place per meaning. When two locations say the same thing, one is duplication.

Failure modes

Use these to diagnose issues the user may be having with the skill.

• Premature completion — ending a step before it's genuinely done, attention slipping to being done. Defence, in order: sharpen the completion criterion first (cheap, local); only if it is irreducibly fuzzy and you observe the rush, hide the post-completion steps by splitting (the sequence cut).
• Duplication — the same meaning in more than one place. Costs maintenance and tokens, and inflates a meaning's prominence on the ladder past its real rank.
• Sediment — stale layers that settle because adding feels safe and removing feels risky. The default fate of any skill without a pruning discipline.
• Sprawl — a skill simply too long, even when every line is live and unique. Hurts readability and maintainability and wastes tokens. The cure is the ladder: disclose reference behind pointers, and split by branch or sequence so each path carries only what it needs.
• No-op — a line the model already obeys by default, so you pay load to say nothing. The test: does it change behaviour versus the default? A weak leading word (be thorough when the agent is already thorough-ish) is a no-op; the fix is a stronger word (relentless), not a different technique.
• Negation — steering by prohibition backfires: don't think of an elephant names the elephant and makes it more available, not less. Prompt the positive — state the target behaviour so the banned one is never spoken; keep a prohibition only as a hard guardrail you can't phrase positively, and even then pair it with what to do instead.

Completion criteria for this skill

When you finish editing or creating a skill, verify:

• Every step ends on a checkable completion criterion.
• No meaning lives in more than one place (single source of truth).
• The description, if present, is pruned to triggers plus reach clauses — no identity that the body already carries.
• The invocation choice (model- vs user-invoked) matches how the skill is actually reached.
• Each leading word is strong enough to change behaviour versus the default.
• No line is a no-op or negation that a positive phrasing could replace.