How to use the Minimal Prose Discipline agent skill

What the skill makes the agent do

The skill applies a writing-time discipline to the prose the agent speaks to a human. Its test for a reply is that every line that survives re-reading is either substance or a sentence the reader needs in order to act.

Code, commands, file paths, exact error messages, numbers, units, API and CLI names and commit-type keywords are protected. They are never reworded. Negators such as not, never, no, only and except are also kept, because dropping one can flip the meaning of a sentence.

• Filler such as just, basically and actually is removed.
• Pleasantries and hedging that add no information are removed.
• Narration of tool calls, such as announcing that the agent is about to search, is removed.
• The conclusion comes first, and the reply ends on the next step.
• Language stays the same. The skill compresses style only, never switches the reply into another language.

What the skill does not cover

The skill governs conversational output only. Artifacts meant for later readers follow other rules: code follows minimal-code-discipline, and documents that humans and agents read follow agent-doc-discipline. Commit messages, doc files, tickets written for humans and memory files are out of scope for this skill, even when the agent writes them.

Install the skill from the ZIP or the Claude apps

Download the ZIP from the skill page. The package contains a single file, SKILL.md. You have two ways to install it.

For Claude Code, unzip the package into your user skills folder at ~/.claude/skills/, or into a project-specific folder at .claude/skills/ inside the project. For the Claude apps, upload the ZIP under Settings → Capabilities → Skills.

• Option A, personal: unzip into ~/.claude/skills/ so the skill is available across your projects.
• Option B, one project: unzip into .claude/skills/ in that project's folder so it applies only there.
• Option C, Claude apps: upload the ZIP under Settings → Capabilities → Skills.

How to trigger the skill with natural requests

The skill description does not list fixed trigger phrases. According to its own description, the agent reaches for the discipline when it is about to write a reply, report or pointer, much as a developer reaches for a coding rule when about to code. So you do not need a special command.

If you want to be certain the style applies, name the skill in your request. Phrasing such as asking the agent to use the minimal-prose-discipline skill for its status updates states your intent in plain words.

• Name the skill directly in a session where you want terse reporting.
• Ask for the result and evidence first when you review many status reports.
• Keep your request plain. The skill is a style rule, so it does not need a long prompt.

Examples of what changes in a reply

Without the skill, a typical report opens with pleasantries and narrates each step. With the skill, the first line states the finding, the protected details appear exactly as produced, and the reply closes with the next action. The example below is an illustration of the style, not output copied from the skill.

• Before: a sentence that says it is happy to help, then narrates a search of the codebase, then hedges about the cause.
• After: the cause and the exact error text in the first line, followed by the next step to take.

What it needs and the safety notes

The skill lists no requirements. It runs no scripts and makes no network calls. It consists of one instruction file, so installing it does not add executable code to your environment.

The skill is built to suspend itself when compression could create risk. Security warnings, irreversible-action confirmations, multi-step sequences where a dropped word could change the order, and passages where the short form is ambiguous are written in full plain prose.

• Requirements: none listed.
• Scripts: none.
• Network use: none.
• Expect full sentences around warnings and multi-step instructions. This is intended behaviour.

Tips for getting the most out of it

Use the skill when you are watching a long session and want short updates. Check the protected details in any reply you act on, because the skill is designed to keep them byte-for-byte, but you should still confirm commands before running them.

Compressed replies omit context by design. If you need an explanation or a teaching tone, ask for it directly or turn the skill off for that session.

• Read the first line for the answer; open the rest only if you need the evidence.
• Ask for the explanation separately when you need it.
• Remember that the skill compresses only the agent's chat output, not the files it writes.