writing

Orchestrates a multi-phase pipeline to draft, critique, and finish long-form prose.

Updated Mar 18, 2026
One-click install
npx skills add https://github.com/pgoell/pgoell-claude-tools --skill writing-pgoell
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: writing
Source: https://github.com/pgoell/pgoell-claude-tools/tree/main/plugins/writing/skills/writing
Command: npx skills add https://github.com/pgoell/pgoell-claude-tools --skill writing-pgoell

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Drafting long-form prose often produces generic, AI-sounding text with weak structure and unearned claims. This Skill runs a structured pipeline that interviews the writer, negotiates an outline, enforces a throughline, drafts, runs a panel of specialized critics, and applies finishing passes so the final piece is specific, rigorous, and in the writer's voice. ## Core Features & Use Cases - Multi-phase pipeline: Interview, outline, throughline gate (10-word compression), draft, panel review, and finishing, with resumable state per project. - Format-aware branching: Narrative formats (essay, blog, talk, newsletter), analytical formats (memo, briefing, announcement) routed through a Pyramid Principle pipeline, and technical formats (tutorial, how-to, reference, explanation) delegated to a tech-doc pipeline. - Critic panel and finishing passes: Seven parallel critics (Hemingway, Hitchcock, Mom reader, Asshole reader, Clarity, Usage, Steel-man, plus Smart-Brevity for memo/newsletter/announcement) and sequential finishing passes (AI-pattern detector, style enforcer, line editor, Sedaris or analytical voice). - Use Case: Ask the agent to write a memo about a Q3 roadmap decision; it runs pyramid intake and construction, drafts directive prose, gets critiqued by eight reviewers, and finishes with an executive-voice pass. ## Quick Start Use the writing skill to draft an essay about why spec-driven development works, starting from an interview and ending with a finished draft.

Frequently Asked Questions about writing

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I write a long-form essay with an AI writing pipeline?

Invoke the writing skill with your topic; it interviews you to extract thesis, lived experience, and counterargument, negotiates an outline, enforces a 10-word throughline, drafts the piece, runs seven critic agents in parallel, and applies four finishing passes. Each phase produces a resumable artifact file.

What formats does the writing skill support?

It supports narrative formats (essay, blog, talk, newsletter), analytical formats (memo, briefing, announcement) routed through a Pyramid Principle pipeline, and technical formats (tutorial, how-to, reference, explanation) delegated to a tech-doc pipeline. Set the format with --format or let it default to essay.

Can I resume an interrupted writing pipeline run?

Yes. The skill records the last completed phase in a state file keyed by working directory. On the next invocation it detects existing artifacts like outline.md or draft.md and offers to resume from the next phase, or you can jump directly with --phase.

How does the critic panel review work?

Seven critics (Hemingway, Hitchcock, Mom reader, Asshole reader, Clarity, Usage, Steel-man) run in parallel, each writing a critique file with a PASS, MINOR, or CRITICAL verdict. Any CRITICAL verdict triggers a draft re-dispatch with feedback, up to two iterations before asking the user.

Does the writing skill work on both Claude Code and Codex?

Yes. It maps capabilities per platform: Agent tool versus spawn_agent for subagent dispatch, TaskCreate versus update_plan for progress, and Read/Write/Edit versus shell and apply_patch equivalents. When subagents are unavailable, phases run inline in the orchestrator.

When should I not use the writing pipeline?

Skip it for short-form content like social posts or quick notes where a six-phase pipeline is overhead. It is designed for long-form prose where structure, critique, and voice finishing justify the multi-phase process.