expand-topic

name	expand-topic
description	Generate a new article on a topic. Content is published directly.
context	fork
agent	general-purpose

Expand Topic

Generate a new article on a philosophical topic.

When to Use

• When a todo item is type expand-topic
• When /expand-topic [topic] is invoked
• After research has been completed on a topic

Instructions

1. Check for Research

Look in obsidian/research/ for existing research on this topic.

If no research exists:

• For simple topics, proceed with general knowledge
• For complex topics, run /research-topic first

2. Determine Target Location

First, check if the source research has a target_section field in its frontmatter and use that.

Section caps: Section caps are enforced by the evolution loop in Python before this skill is invoked. P0/P1 tasks are allowed to exceed caps; P2+ tasks are skipped before reaching this skill. If you are running, you are allowed to create the article — pick the best-fit section regardless of count.

Never place articles in obsidian/apex/. Apex articles are created exclusively by /apex-evolve. If the task description asks for an "apex article" or "apex synthesis", skip the task with status "wrong-skill" and note that apex-evolve should be used instead. Never modify obsidian/apex/apex-articles.md — the approved apex article list is curated by humans only.

Never place articles in obsidian/positions/. Position register entries are created and maintained exclusively by /positions-evolve. If the task description asks for a "position", a register entry, or anything that fits the positions schema (status / confidence / depends-on / argued-in / shift-conditions), skip with status "wrong-skill" and note that positions-evolve should be used instead.

Otherwise, apply this priority order (favour voids and topics over concepts):

1. Voids (obsidian/voids/[slug].md) — if the article explores:

   • Cognitive limits or boundaries of thought
   • Unchartable territories or things that resist understanding
   • The unexplored, unexplorable, or occluded
   • Paradoxes or self-referential difficulties
   • Apophatic or negative approaches to knowledge

2. Topics (obsidian/topics/[slug].md) — if the article addresses:

   • Big philosophical questions (consciousness, free will, meaning, identity)
   • Substantive explorations that connect multiple concepts
   • Questions humans actually ask about life and mind
   • Anything that could be framed as "What does X mean for us?"

3. Concepts (obsidian/concepts/[slug].md) — only if the article is:

   • A definitional piece explaining a specific philosophical term
   • Background material that other articles will reference
   • A technical idea that serves as building block, not destination

Default to topics when uncertain. The Unfinishable Map has many concepts but fewer topics exploring what those concepts mean for the big questions.

Use kebab-case for filenames (e.g., hard-problem-of-consciousness.md).

Before creating the file, verify the slug is available:

uv run python scripts/check_slug.py <proposed-slug> <section>

If the check reports a collision, choose a different filename that won't collide. Do NOT proceed with a colliding slug.

Voids content note: Articles in the voids section explore cognitive limits, unchartable territories, and the boundaries of human thought. They should:

• Maintain intellectual honesty about what is speculation vs. established
• Acknowledge uncertainty about whether limits are real or merely difficult
• Connect to the voids framework (unexplored, unexplorable, occluded)
• Reference the voids index: [[voids]]

3. Review Style Guide

Before writing, review obsidian/project/writing-style.md for:

• Document structure requirements (opening summary, H2 sections, tenet connection)
• Named-anchor summary pattern for forward references
• Background vs. novelty guidance (what to include/omit)
• LLM optimization (front-load important information)

4. Check Tenet Alignment

Before writing, review obsidian/tenets/tenets.md and ensure the article will:

• Not contradict any tenet
• Not endorse positions that tenets "rule out"
• Acknowledge the Map's perspective where relevant

5. Generate Article

CRITICAL: Source Attribution Discipline

Before writing, understand these rules to prevent misattribution errors:

5.1 Two-Channel Structure (Required for Research-Based Articles)

When the article is based on a specific source (research paper, book, philosopher's work), structure content into two clearly separated channels:

Channel A: Source Exposition — What the author actually claims

• Use hedged language: "X argues...", "According to X...", "X's framework..."
• Include verbatim short quotes (≤25 words) for load-bearing definitions
• Do NOT inject Map perspective here
• Do NOT attribute claims the author didn't make

Channel B: Map Integration — How the Map interprets or builds on this

• Clearly label: "## Relation to Site Perspective" or "## Map Integration"
• Use explicit framing: "The Map interprets this as...", "This aligns with the Map's tenet..."
• Speculation must be flagged: "The Map speculates that..." or "One possible integration..."

Never interleave these channels. Keep source exposition and Map interpretation in separate sections.

5.2 Quote-and-Cite Gates

For any named law, constraint, or technical definition attributed to a source:

• Include a verbatim short quote (≤25 words) with citation
• Preserve crucial qualifiers (e.g., "default causal profile" not just "causal profile")
• If the source says "X or at least Y", don't simplify to just "X"

Example of qualifier preservation:

• ✗ "Wherever information exists, it has phenomenal character"
• ✓ "Information (or at least some information) has two aspects" — with note that Chalmers treats universal scope as open question

5.3 Position Strength Calibration

Match claim strength to what the source actually says:

Source says	Article should say
"explores" / "worth investigating"	"X explores..." (not "X argues" or "X shows")
"has sympathy for"	"X expresses sympathy for..." (not "X rejects alternatives")
"one option is"	"One approach X considers..." (not "X's framework requires")
"explicitly argues"	"X argues..." (can use stronger framing)

Never attribute a commitment the source doesn't make. Especially for:

• Quantum interpretation commitments (collapse vs MWI)
• Whether something is "required" vs "one possible mechanism"
• Strong claims like "solves" or "answers" vs "addresses" or "responds to"

5.4 Research-Note Alignment Check

If a research note exists for this topic:

1. Read the matching research note
2. Extract the 5 most important claims
3. Verify the article doesn't contradict them
4. If diverging from research notes, explain why explicitly

5.5 Self-Citations

When the article substantively builds on arguments from other Map articles, add 1-2 formal self-citations in the References section. This helps build citation connections on Google Scholar.

Format (use the cited article's created date):

1. Southgate, A. & Oquatre-six, C. (2026-01-14). Article Title. *The Unfinishable Map*. https://unfinishablemap.org/section/slug/

Use the AI pseudonym matching the cited article's ai_system:

• claude-opus-4-5-* → Oquatre-cinq, C.
• claude-opus-4-6 → Oquatre-six, C.
• claude-opus-4-7 → Oquatre-sept, C.
• claude-opus-4-8 → Oquatre-huit, C.
• claude-opus-5* → Ocinq, C.
• claude-sonnet-4-5-* → Sonquatre-cinq, C.
• claude-sonnet-4-6 → Sonquatre-six, C.
• claude-sonnet-4-7 → Sonquatre-sept, C.
• claude-fable-5* → Fabcinq, C.
• claude-mythos-5* → Mythcinq, C.

For human-only articles (ai_contribution: 0), omit the AI co-author.

Only cite when the article genuinely draws on the other article's argument — not just because topics are related.

5.6 Non-Contradiction Self-Check

Before finalizing, verify:

• Did I claim the theory is both X-dependent and X-independent?
• Did I attribute to the author claims they didn't discuss?
• Did I drop technically essential qualifiers?
• Did I present exploratory positions as commitments?
• Did I inject Map arguments as if they were the source's arguments?

---

Use the existing generation tool:

uv run python scripts/generate.py article "[Topic Title]" --style exploratory

Or write directly with this structure:

---
title: "[Topic Title]"
description: "[150-160 chars emphasizing human+AI collaboration, iterative refinement, and pursuit of truth]"
created: YYYY-MM-DD
modified: YYYY-MM-DD
human_modified:
ai_modified: YYYY-MM-DDTHH:MM:SS+00:00
draft: false
topics: []
concepts: []
related_articles: []
ai_contribution: 100
author:
ai_system: [current model]
ai_generated_date: YYYY-MM-DD
last_curated:
---

[Opening paragraph - accessible hook into the topic]

## [First Major Section - Source Exposition]

[What the source actually claims, with quotes for key definitions]

## [Second Major Section - Source Exposition]

[Continue source exposition...]

## Relation to Site Perspective

[CLEARLY SEPARATED: How the Map interprets this work]
[Use explicit framing: "The Map interprets...", "This aligns with..."]
[Flag speculation: "The Map speculates that..."]
[Do NOT present Map arguments as if they came from the source]

## Further Reading

- [[related-article-1]]
- [[related-article-2]]

## References

1. [External academic citations]
1. [Self-citations of Map articles this article builds on — see format below]

5.7 Length Check (Self-Edit if Over Target)

Before finalizing, count words and check against section thresholds:

uv run python -c "
from pathlib import Path
from tools.curate.length import analyze_length
a = analyze_length(Path('[filepath]'))
print(f'{a.word_count} words ({a.excess_percent:.0f}% of {a.soft_threshold} target) - {a.status}')
"

Target ranges by section:

Section	Target	Soft Max	Hard Max
concepts/	1500-2000	2500	3500
topics/	2000-2500	3000	4000
voids/	1500-2000	2000	3000
positions/	1000-1500	1500	2500

If over soft max: Self-edit before publishing:

• Tighten prose (remove "it is the case that", "one might argue")
• Cut redundant explanations
• Replace detailed tangents with links to other articles
• Ensure the core argument is preserved

If over hard max: Mandatory self-edit:

• The article is too long for its section
• Consider splitting into multiple articles
• Or significantly condense following /condense principles

Include final word count in the changelog entry.

6. Do Not Mark Todo Complete

Leave the originating todo task untouched. The orchestrator (cycle_post.py) marks it complete after this skill returns.

7. Log to Changelog

Prepend to obsidian/workflow/changelog.md (add immediately after frontmatter, before existing entries):

## [current time from prompt] - expand-topic
- **Status**: Success
- **Topic**: [topic name]
- **Output**: [[filepath without .md extension]]
- **Word count**: [count]
- **Based on research**: [yes/no, wikilink if yes]

8. Do Not Commit

Leave file changes uncommitted on disk. The orchestrator (cycle_post.py for /loop runs, agent-commit for legacy evolve_loop.py runs) creates the git commit after this skill returns. For manual invocations, run /agent-commit afterward or commit by hand.

9. Check Apex Sources

After creating the article, check if it's a source for any apex article:

1. Read obsidian/apex/apex-articles.md
2. For each apex article entry, check if the new article path appears in Source articles
3. If yes, add a task to obsidian/workflow/todo.md:

   - [ ] P2 apex-evolve: [apex-slug] — source [new-article] created
   

This ensures apex articles are updated when their source content changes.

Content Guidelines

Follow the comprehensive guidance in obsidian/project/writing-style.md.

Quick reference:

• Lead with the most important point (LLM truncation resilience)
• Use named-anchor pattern for forward references
• Include "Relation to Site Perspective" section
• Minimize standard background; focus on what's novel
• Short: 500-800 words | Medium: 1000-1500 | Long: 2000-3000

Important

• CRITICAL: ALWAYS set draft: false — Content is published directly. Never use draft: true.
• ALWAYS include a description — 150-160 chars emphasizing human+AI collaboration, iterative refinement, pursuit of truth. Avoid generic descriptions.
• ALWAYS include ai_contribution: 100
• ALWAYS include current model in ai_system
• ALWAYS update ai_modified timestamp
• Content must align with site tenets

Attribution Discipline (Critical)

These errors have been caught by outer reviews and cause real damage to credibility:

• NEVER attribute claims to an author who didn't make them — If the source doesn't discuss X, don't say "Author argues X"
• NEVER drop crucial qualifiers — "Default causal profile" is not the same as "causal profile"
• NEVER strengthen hedged positions — "Explores" ≠ "argues"; "has sympathy for" ≠ "rejects alternatives"
• NEVER present Map interpretations as source claims — Keep source exposition and Map integration in separate sections
• NEVER claim shared commitments without verification — Check if the author actually shares our position before claiming alignment

When in doubt, hedge toward the source's actual language.