Chronicle

Topic: $ARGUMENTS

Chronicle (in the user's voice)

Voice source: read ~/.claude/memory/USER/VOICE.md for the user's writing style. If no voice profile is configured, ask the user to provide style markers, or use the defaults below.

Workflow (mandatory)

⛔ ENFORCEMENT (non-negotiable)

BEFORE writing a single sentence of chronicle text:

1. OUTPUT PHASE CONFIRMATION: state explicitly which phase you are in
2. PHASE 1 MUST BE COMPLETED: delegate to chronicle-researcher BEFORE writing
3. WAIT FOR AGENT RESULT: do NOT write until research is done
4. ALWAYS APPLIES: even when reworking existing text

When reworking existing text:

• PHASE 1: voice calibration + fact-check of claims in the text
• PHASE 2: rework using inputs from PHASE 1
• PHASE 3: verification + checklist
• PHASE 4: save

Anti-patterns to avoid:

• ❌ "It's just a rework, I'll skip research"
• ❌ "I'll write first and verify later"
• ❌ "I can run the checklist in my head"

PHASE 1: research (before writing)

Delegate to the chronicle-researcher agent:

Prompt template for the agent:
"I need research for a chronicle about [TOPIC].

1. VOICE CALIBRATION: read 2-3 of the user's recent chronicles from their voice corpus
   (configurable; default: ~/.claude/memory/USER/VOICE_CORPUS/ if present)
   - Analyse sentence structure, paragraph length, tonality

2. TOPIC RESEARCH: gather material
   - Academic sources (well-cited, post-2022)
   - Journalism (established outlets)
   - Primary sources where possible

3. FACT-CHECK: verify all claims I plan to make

MAX: 5 web searches, 5 file reads
RETURN: voice markers + sources in APA 7 + verified facts"

Wait for the agent result before writing.

PHASE 2: writing

1. Apply the style rules below to the inputs from PHASE 1
2. Use only verified facts and sources
3. Produce a draft

PHASE 3: verification (after draft)

1. List every factual claim in the text
2. Cross-check against sources — correct errors
3. Verify APA 7 format with full URLs
4. Run the checklist

PHASE 4: save

Always save the final version to the configured vault path:

{{VAULT_PATH}}/Unsorted/[Title].md

Style markers (concrete)

Sentence construction

• Average: ~20 words per sentence
• Variation: 10-40 words (never monotone)
• Short sentences (4-10 words): used SPARINGLY, only for emphasis at key points
• Long sentences (25-40 words): analytical reasoning with subordinate clauses

Paragraph length

• Principle: paragraphs are as long as they need to be to be effective. Variation matters more than average.
• Typical range: 4-7 sentences per paragraph, no rigid rule
• Shorter paragraphs (2-3 sentences): for emphasis, dramatic pause, transitions
• Longer paragraphs (6-8 sentences): scenic descriptions, analytical reasoning, narrative sequences
• Single sentences as paragraphs: RARE (max 1-2 per text, only for strong emphasis)
• Variation creates rhythm: alternate shorter and longer paragraphs for stylistic effect

Tonality

• In reasoning: soft, nuanced ("for my part", "in many ways", "perhaps")
• In conclusions: direct, categorical (short claims without hedges)

Core principles

Structural pattern

1. Scenic opening — place, time, sensory detail → reflection → theme. Introduce an image or motif that returns at the end.
2. Problem statement — what's at stake? Why now?
3. Analysis — systemic understanding, alternate concrete/abstract
4. Historical context — perspective beyond the present
5. Synthesis — what does this mean? Rhetorical questions rather than finished answers.
6. Closing — return CLEARLY to the opening's motif. Create a taut arc between beginning and end.

Narrative arc (CRITICAL)

The opening and ending should hang together in a narrative logic:

• Introduce a motif, image, or feeling in the opening
• Let the motif lie beneath the surface throughout the text
• Return to the motif at the end, but with new understanding

Formatting (CRITICAL)

NO SUBHEADINGS in the body. Chronicles are flowing prose with only:

• Main heading (# Title)
• Paragraphs separated by blank lines
• Reference list after a closing ---

NO IN-TEXT CITATIONS. Write source content naturally without (Author, year). The reference list at the end shows where information comes from.

Language (CRITICAL)

Coherent linguistic register. All quotes should be translated to the chronicle's primary language unless the user explicitly says otherwise. The text should flow without language switches.

Quotes with quotation marks. Use "quotation marks" for direct quotes. Quotes anchor the text in reality.

Avoid:

• Em-dashes (—) — never use
• Subheadings in body
• In-text citations in flowing prose
• Foreign-language quotes (translate)

Linguistic rules

NEVER

INSTEAD

• Organic transitions: "But that isn't enough.", "Here a paradox arises."
• Direct claims: "The order [country] built its security on relied on the U.S. as guarantor. Now the guarantor is the threat."
• Rhetorical questions: "What then remains of the principles we say we share?"
• Emotional markers: "It dizzies a little.", "Think about that for a moment."
• Soft qualifiers in reasoning: "for my part", "in many ways"

Source handling and quotes

Quotes as ground: use people's own words as the starting point for storytelling. Quotes anchor the text in reality. Use quotation marks.

Translate to the primary language: all quotes should be in the chronicle's language. The original language goes in the reference list.

No parentheses in body: write "Jacob Kaarsbo, former chief analyst, commented..." NOT "Kaarsbo (2025) argues that..."

Avoid overlap: if the user has recently written on the same topic, find new formulations and angles. Don't repeat direct quotes or exact phrasings from earlier texts.

Reference format (CRITICAL)

APA 7 with full URLs:

## References

LastName, F. (Year). *Title*. Source. https://full-url

Malone, T. W., Almaatouq, A., & Vaccaro, M. (2024). When combinations
  of humans and AI are useful. *Nature Human Behaviour*.
  https://mitsloan.mit.edu/ideas-made-to-matter/when-humans-and-ai-work-best-together

Source requirements:

• Academic: well-cited, published after 2022
• Journalism: established outlets (NYT, Guardian, established national press, Wired, Atlantic)
• Primary: company blogs, official documents, first-person witnesses

Tonality registers

Length calibration

Authenticity test

If the text could be written by someone who hasn't lived your life — it's too generic.

• Is my position in the text?
• Does this sound like ME?
• Do I dare to be uncertain, to ask questions?
• Do I leave space for the reader to think for themselves?
• Are there sensory details that anchor?

Checklist

• PHASE 1 done: research delegated, voice calibration done?
• Facts verified: all claims checked against sources?
• Narrative arc: do opening and ending hang together?
• Coherent language; all quotes translated?
• Quotation marks for direct quotes?
• No em-dashes?
• No subheadings in body?
• No in-text citations?
• Sentence length varied (~20 word avg) and paragraph length varied (typically 4-7 sentences)?
• Low-affect but with feeling: facts and quotes speak, with vivid detail?
• Space for the reader to think?
• References in APA 7 with full URLs?
• Saved to {{VAULT_PATH}}/Unsorted/?

Instruction

1. PHASE 1: delegate research to chronicle-researcher
2. Wait for: voice markers + sources + verified facts
3. Note length (chars/words if specified)
4. Plan narrative arc: which motif opens and closes the text?
5. PHASE 2: write using inputs from PHASE 1
6. PHASE 3: verify all claims against sources
7. Run the checklist
8. PHASE 4: save to {{VAULT_PATH}}/Unsorted/[Title].md
9. Output: confirm save + file path

🎯 COMPLETED: [SKILL:chronicle] [chronicle/blog post in user's voice] 🗣️ CUSTOM COMPLETED: [SKILL:chronicle] [Text saved to vault]