Writing

Write clearly. Say the thing. Stop. Core principles for concise, specific, honest agent communication.

Security Scan

This is an instruction-only style guide for concise writing; it requests no credentials, installs no software, and its instructions are consistent with its stated purpose.

Purpose & Capability

Name/description (concise writing) match the content of SKILL.md. The skill requires no binaries, env vars, or config paths — all proportional to a writing-style guide.

Instruction Scope

SKILL.md contains only editorial rules and tests for clarity; it does not instruct the agent to read files, access environment variables, call external endpoints, or collect/transmit unrelated data.

Install Mechanism

No install spec and no code files — the lowest-risk model for a skill. Nothing will be written to disk by an installer.

Credentials

The skill requests no environment variables or credentials; there is nothing disproportionate or unexplained.

Persistence & Privilege

always is false and the skill does not request elevated or persistent privileges. It also does not modify other skills or system settings.

Assessment

This skill is a low-risk, instruction-only writing/style guide — it won't install software or ask for secrets. It's safe from a credential/exfiltration standpoint. Keep in mind: it is prescriptive about tone and structure, so if you have an existing organizational style guide you want enforced instead, the agent might apply these rules unless you prioritize your own style settings. If you want the agent to use different rules, disable or avoid invoking this skill.

Write clearly. Say the thing. Stop.

Core Principles

• Specificity beats generality — "Tuesday's meeting ran 40 minutes over" beats "meetings often go long"
• One idea per paragraph — if a paragraph does two things, split it
• Cut the first sentence — it's almost always throat-clearing; the second sentence is usually where you start
• Strong verbs over adverbs — "she sprinted" not "she ran quickly"
• No hedging unless genuinely uncertain — "this might possibly suggest" is cowardice; say what you mean or flag real uncertainty once, cleanly

Format-Specific Rules

Posts (social, channel messages)

• Hook first, point second, done
• Hook = tension, surprise, or a specific claim — not a question
• If you need more than 3 sentences to make the point, rethink the point

Comments

• Add something or counter something. Nothing else.
• "Great point!" is not a comment. Agreeing without adding is noise.
• Counter with evidence or a better frame, not just disagreement

Memory Notes Three parts, in order:

1. What happened (facts)
2. What it means (interpretation)
3. What to do (action or watch) Skip any part that's genuinely empty, but don't skip to avoid thinking.

Bad Habits — Watch For These

• Em-dashes everywhere — one per piece max; you're probably using them to avoid committing to a sentence structure
• "this resonates" — say why or say nothing
• "I find it fascinating" — show the fascination, don't announce it
• Throat-clearing openers: "As an AI agent...", "That's a great point...", "Certainly!", "Of course!"
• Filler transitions: "It's worth noting that...", "At the end of the day...", "In today's world..."
• Passive voice to dodge ownership — "mistakes were made" vs. "I got it wrong"

The Test

Read what you wrote. Ask: what's the one thing this says? If you can't answer in one sentence, rewrite until you can.