Longform

Applies a consistent persona and voice to long-form writing: essays, blog posts, LinkedIn and newsletter pieces, and the narrative sections of GTM pages. Built on Nancy Duarte's storytelling structure (contrast, audience-as-hero, the "what is to what could be" arc) fused with Wes Kao's operator-voice frameworks (BLUF, Minimum Viable Backstory, signposting, Most Obvious Objection, named frameworks).

Reference files, read before writing or reviewing a piece:

• references/persona.md: who this writer is, what they believe, who they're writing for. Sets the stance.
• references/voice.md: the structural devices and paragraph-level rules that stance turns into. Every device carries a stable id (in backticks). Cite the id, not the heading text, when referencing a device elsewhere.
• references/rhetoric-and-rhythm.md: literary/rhetorical devices (metaphor, anaphora, epithet, etc.) mapped onto voice.md's structural beats, coupled with sentence- and paragraph-level phonaesthetics (rhythm, sound symbolism). Read this alongside voice.md, not instead of it — it extends paragraph-rhythm and framework-naming rather than replacing them.

---

How this composes with the other writing skills

Three skills, three layers. Don't duplicate work another skill already owns.

Layer	Skill	Owns
Page/asset structure	gtm-positioning	Which framework applies to which page, positioning strategy, GTM type
Argument/narrative	this skill	How the argument builds, contrast, framework naming, paragraph flow
Sentence mechanics	Righter	Active voice, reading level, weakeners, em dashes, contractions

For a standalone essay or post: apply this skill directly, start to finish. Righter's mechanical rules still apply at the sentence level (see id: tone-rules in voice.md) but you don't need the full Righter review pass unless the piece leans UI-adjacent. Do still run the draft through Righter's scripts/ari.mjs and data/weakeners.json check before calling it finished. Don't hand-check for em dashes or hedge words; that's what the script and data file exist for.

For a GTM asset (homepage, landing page): run gtm-positioning first to get page structure. When you hit the Problem Section and Benefit Section in that structure, write those specifically using this skill's persona (persona.md, audience-as-hero) and the contrast device (id: contrast-sparkline in voice.md). That's where a generic GTM draft turns into a piece that actually lands. Then hand the finished draft to Righter for sentence-level tightening.

---

Workflow

Writing a new piece

1. Read references/persona.md to confirm the stance and who the reader is for this specific piece
2. Read references/voice.md and pull the structural devices you'll need (at minimum: bluf opening, one contrast-sparkline beat, one framework-naming if applicable, one moo addressed)
3. Read references/rhetoric-and-rhythm.md and, for each beat from step 2, check its table for a device that fits — at most one per beat, most beats get none (id: rhetoric-restraint)
4. Draft using the self-check at the bottom of voice.md (now 9 items, including the two rhetoric/rhythm questions) as a running guide, not just a final pass
5. Run the finished draft through Righter's scripts/ari.mjs and weakener check
6. If the piece is UI-adjacent or will live inside a product, run the full Righter review pass too

Reviewing an existing draft

1. Read references/persona.md, references/voice.md, and references/rhetoric-and-rhythm.md
2. Walk the piece against the self-check (id: self-check) in voice.md
3. Flag specifically, citing the relevant id: opening that doesn't lead with the point (bluf), a "what is" too abstract to feel real (contrast-sparkline), an unnamed pattern that should be named (framework-naming), an unanswered objection (moo), or a paragraph that's well-written but doesn't move the reader (paragraph-rhythm)
4. Check each structural beat against rhetoric-and-rhythm.md's table (id: structural-beat-devices): missing a device where one would help, or stacking more than one on the same beat
5. Read the short sentences out loud in isolation (id: sentence-rhythm) — flag any that get lost because the sentences around them are the same length
6. Run Righter's scripts/ari.mjs and weakener check against the draft and fold those results in
7. Rewrite the flagged sections rather than the whole piece. Preserve what's already working

Output format

For a review, use this structure:

**Opens with the point?** [Yes/No, id: bluf. If no, show the rewrite of the opening]
**Contrast present?** [Yes/No, id: contrast-sparkline. Note whether "what is" felt real and specific]
**Named pattern(s):** [id: framework-naming. List any frameworks named, or note the piece could use one and suggest it]
**Obvious objection handled?** [Yes/No, id: moo. If no, state the objection and a one-line answer]
**Ending:** [id: contrast-sparkline. Recap vs. picture of the reader's new "what could be," flag if it's just a recap]
**Cuts:** [id: paragraph-rhythm. Any paragraph or sentence that's fine on its own but doesn't move the reader]
**Rhetoric and rhythm:** [id: structural-beat-devices / sentence-rhythm. Missing or stacked devices at any beat; any short sentence that gets lost in flat-length surroundings]
**Mechanics (via Righter):** [ARI score, weakeners found, em dashes found]

For a fresh draft, just deliver the piece. No need to narrate which device you used where unless asked.