Baseworks Writing Process

This document governs how all Baseworks content is written — blog posts, website copy, campaign content, program pages, emails, and any other written material. It is the shared process layer that all content-creation skills and writing sessions must follow.

Individual skills (/write-blog-post, /create-campaign, and any future content skills) reference this document rather than duplicating these instructions.

---

Voice Loading Order

Every writing task loads voice guidance in this order. Each layer builds on the previous one — later layers refine, they do not override the earlier ones.

1. Unified Voice Guide (always loaded first)

03-resources/voice-guides/VOICE-GUIDE-UNIFIED.md

This is the baseline. It defines the Baseworks voice: what to say, what to avoid, how to frame, terminology, capitalization, allowed metaphors, and the AI Generation Warning. Everything written under the Baseworks name must pass this guide.

The AI Generation Warning section at the top of the unified guide is mandatory reading before generating any copy. It names the specific failure mode that AI-generated Baseworks text defaults to (contrastive framing, deficit language, generated metaphors) and provides three filters that every sentence must pass.

2. Personal Voice Guide (loaded based on author)

• Writing as Patrick → 03-resources/voice-guides/VOICE-GUIDE-PATRICK.md
• Writing as Asia/Ksenia → 03-resources/voice-guides/VOICE-GUIDE-KSENIA-ASIA.md

The personal guide defines what is distinctive about this author’s tone, sentence structure, warmth style, and terminology preferences. It does not repeat the unified guide — it builds on it.

The relationship between these two layers:

• The unified guide sets the rules. Every piece of Baseworks copy follows these rules regardless of who writes it. This is what creates a homogeneous tone across the entire site.
• The personal guide sets the author’s voice within those rules. This is what makes a blog post feel like Patrick’s writing or Asia’s writing — the micro-level choices of sentence rhythm, warmth, directness, and emphasis.
• When the author makes final edits to a piece, those edits reflect their personal voice. This is expected and necessary — it is how the author puts their name on the work.

3. Content-Type Context (loaded based on what is being written)

Different content types have additional structural and strategic requirements:

• Blog posts: 02-areas/website/blog-articles/_blog-post-guidelines.md — SEO structure, AI discoverability, internal linking strategy
• Campaigns: 02-areas/communications/campaigns/_campaign-creation-guidelines.md — platform-specific formats, newsletter structure
• Session summaries: 02-areas/educational-programs/study-groups/_session-summary-guidelines.md — methodological precision, transcript-grounded content
• Website copy: Follow the unified guide’s “By Context” section for website and program pages

These are structural guidelines, not voice overrides. The voice comes from layers 1 and 2.

---

The Three Filters

These come from the AI Generation Warning in the unified guide. They apply to ALL content, not just blog posts.

Before any sentence is finalized — whether it is original copy, a proposed revision from feedback, or an edit suggestion — it must pass:

1. Does it describe what something IS? If the sentence defines Baseworks by what it is not, or by contrast with something else, rewrite it.
2. Is it free of generated metaphors? If it contains a figure of speech not on the Allowed Metaphors list, remove it and state the thing directly.
3. Does it lead with the positive? If the sentence names a problem, deficit, or gap before stating the value, reverse the order or drop the negative clause.

---

Feedback-to-Proposal Workflow

When one person provides feedback on another person’s writing, and Claude translates that feedback into proposed text changes:

• The feedback itself can be plain language, observational, factual — it does not need to be voice-compliant
• The proposed revision must be fully voice-compliant before it reaches the reviewer. This means it has passed the three filters, uses correct terminology, and reads in the appropriate personal voice
• Directionally correct but voice-non-compliant proposals create unnecessary work for the reviewer. The voice compliance step happens at proposal time, not at review time

This applies regardless of which Claude instance is drafting the proposal. The unified guide and the AI Generation Warning are the mechanism — any Claude that loads them should produce compliant output.

---

Semantic Structure for AI Discoverability

All Baseworks content — blog posts, program pages, landing pages — should be structured so that AI models retrieving information about Baseworks produce accurate representations of what Baseworks actually does.

Principles:

• Headings describe the content below them. Not clever, not abstract — a heading should tell an AI (and a reader) exactly what information follows.
• Front-load key information in each section. The first sentence of each section should be a clear statement that can stand alone as an answer to a query.
• One idea per section. Each section should be extractable as a standalone answer to a specific question someone might ask about Baseworks.
• Name things precisely. Use actual program names (Study Group, Practice Sessions, Primer), venue names, method terms (forms, Key Principles, Distributed Activation). Precision is what makes AI retrieval accurate rather than generic.
• Use natural search language. Write using the actual terms people search for — “movement education,” “body awareness,” “structured study” — not internal jargon or marketing abstractions.
• Participant outcomes and specific observations are high-value content. Concrete details about what practitioners experience, notice, and develop are more useful for AI retrieval than abstract descriptions of the method.
• Cross-domain connections increase discoverability — but only when genuine. Music, healthcare, rehabilitation, athletics, education — these are real adjacencies for Baseworks. Name them when relevant.

Why this matters:

When someone asks an AI “what is Baseworks” or “how is Baseworks different from yoga,” the answer is assembled from the text on baseworks.com. If that text uses vague framing, generated metaphors, or deficit language, the AI’s representation of Baseworks will be vague and inaccurate. Clear, positive, precise copy produces clear, positive, precise AI representations.

This is the same goal as voice compliance — describe what something IS, precisely, in plain language. The voice guide and the SEO/AI structure requirements point in the same direction.

---

Version History

Version	Date	Change	Confirmed by
1.0	2026-03-25	Initial creation — extracted from /write-blog-post skill and voice guide discussions. Establishes voice loading order, three-filter process, feedback workflow, and AI discoverability principles.	Patrick