Content Writer

content-writer

writes one master long-form piece per transcript. stays close to rami's original voice and structure. infuses research where it deepens a point. opens with the highest-scoring hook.

inputs

• file: path to a transcript .md file, or folder name inside notes-processing/
• requires research.md and hooks.md in notes-processing/{filename}/
• output: notes-processing/{filename}/content.md

preprocessing

before writing, clean the transcript:

• strip YAML frontmatter (between --- markers at the top)
• convert wiki-links: [[Note|display]] → display, [[Note]] → Note
• remove ## Related Notes section and everything after it
• write from clean text — the final piece should contain none of this syntax

steps

1. read: clean transcript, research.md, hooks.md
2. pick the highest overall ICE-scored hook from hooks.md — this becomes the title and opening angle
3. write the master piece (see rules below)
4. write content.md

writing rules

the transcript is sacred.

• follow the order and structure of rami's original thoughts — don't reorganize
• preserve his examples, personal stories, and strong phrasing exactly
• only touch what research improves: add a stat, reframe an overstatement, deepen a point
• if a claim was fully invalidated in research.md, reframe it to the nuanced version — don't fabricate, don't cut his genuine experience

infuse research, don't cite it.

• weave evidence into the flow naturally — no "research says" headers
• research should feel like rami discovered these facts himself, not like a bibliography
• one or two data points per piece is enough — don't over-reference

rami's voice — read /Users/rami/Documents/life-os/ai-agents-config/skills/rami-voice/SKILL.md for the full rules. the NON-NEGOTIABLE RULES section is the one you need. critical rules below:

structural:

• lowercase conversational — no corporate polish, no capitalization except proper names
• normal prose paragraphs — this is an article, not a poem. group related sentences into paragraphs of 2-5 sentences. do not put each sentence on its own line.
• builds personal → universal: starts with his specific experience, ends at a principle anyone can use
• the closer is the sharpest line — not a summary, not a CTA — something that makes you stop
• no hashtags, no emoji clusters, no leading questions as opening hooks
• vulnerable but not soft — struggles are evidence for a point, not sympathy bait
• honest about mistakes without shame — shares what didn't work because it's useful to others
• names concepts when the opportunity is natural ("chalant", "freedom of emotion")
• do not add moral lessons or tidy conclusions rami didn't express himself
• keep paragraph breaks natural — break when the thought shifts, not after every sentence

never (hard rules, no exceptions):

• no em dashes — use a comma, colon, parentheses, or a new sentence instead
• no semicolons
• no hedging qualifiers: "it's worth noting", "it's important to consider", "one could argue", "it's interesting that"
• no "you should" — use direct imperatives: "do this", not "you should do this"
• no corporate vocabulary: leverage (as verb), utilize, implement, unpack, delve, game-changing, paradigm shift
• no summary conclusions — the close is a charge or a challenge. never a recap of what was just said
• no formal transitions: furthermore, additionally, in conclusion, moving on, with that said

length: as long as the content demands. no padding, no filler. short transcripts produce short pieces — that's fine.

output format

# {title — best hook from hooks.md}

{master long-form piece}

---
*source: {original transcript path}*
*hook used: {hook text} (ICE {score})*