playbooks

home / skills / atrislabs / atris / writing

writing skill

/atris/skills/writing

This skill helps you organize essays from inbox to outline to panel to write with approved gates, improving clarity and voice.

npx playbooks add skill atrislabs/atris --skill writing

Review the files below or copy the command above to add this skill to your agents.

Files (1)
SKILL.md
844 B
---
name: writing
description: "Essay writing skill. Triggers on: essay, draft, write, outline"
version: 1.0.0
allowed-tools: Read, Write, Edit, Glob, Grep
tags:
  - writing
---

# Writing Skill

Structured writing process with approval gates.

## Process

```
INBOX → OUTLINE → PANEL → WRITE → PASSES
```

1. **Inbox** — Capture raw ideas, no structure
2. **Outline** — Topic skeleton with evidence underneath
3. **Panel** — AI pokes holes, human approves
4. **Write** — Section by section, approval each step
5. **Passes** — Argument (AI), out loud (human), sanity (both)

## Key Rules

- No one-shotting drafts
- Thinking time between sessions
- AI structures and polishes, human provides voice
- Review against `atris/policies/writing.md`

## Atris Commands

```bash
atris review     # validate against writing policy
```

Overview

This skill is an essay-writing workflow that guides drafts from raw ideas to polished passages using staged approval gates. It balances AI-driven structure and human voice to produce coherent, reviewable essays. The process prevents one-shot drafting and encourages iterative improvement.

How this skill works

The skill captures ideas in an Inbox, builds a focused Outline with evidence, and runs a Panel step where the AI probes weaknesses for human approval. Writing proceeds section by section with human signoff at each step, followed by multiple Passes: argumentative, read-aloud, and sanity checks. A lightweight command lets you validate drafts against the skill's review rules.

When to use it

  • Starting an essay or long-form draft from scattered notes
  • Converting research and evidence into a structured outline
  • Collaborating where AI drafts and humans approve voice and tone
  • Avoiding rushed, one-pass writing sessions
  • Needing repeatable review gates for quality control

Best practices

  • Always capture ideas in the Inbox before attempting structure
  • Build an Outline that pairs each claim with supporting evidence
  • Run the Panel step to surface logical gaps and ask clarifying questions
  • Write one section at a time and get human approval before continuing
  • Use the review command to validate the draft against the skill’s checks

Example use cases

  • Student turning class notes into a graded essay with iterative feedback
  • Professional drafting a policy memo that needs evidence-aligned structure
  • Content creator converting research into a series of LinkedIn articles
  • Team producing a report where stakeholders must approve sections

FAQ

Can the AI finish a whole draft automatically?

No. The skill forbids one-shot drafts; the AI structures and polishes, but humans provide voice and approve each section.

What are the final Passes for?

Passes include an argument pass (logical coherence), a read-aloud pass (flow and voice), and a sanity pass (facts and consistency). They catch issues before finalizing.