Enhance writing-fragments skill to capture initial user prompts as fragments

skills/in-progress/writing-beats/SKILL.md

@@ -11,7 +11,7 @@ If the user did not say where to save the article, ask once and remember the pat
 
 Then run a beat-by-beat journey:
 
-1. Write 2–3 candidate **starting beats**, drawn from the raw material. Each is a different entry point into the article. Show the user the beats before writing it to the article file. The user picks one.
+1. Write 2–3 candidate **starting beats**, drawn from the raw material. Each is a different entry point into the article. Show the user the beats before writing it to the article file. The user picks one. Preview what beats that might lead to once written - as if the user is seeing a little way down the path.
 2. Once the user picks a starting beat, write **only that beat** to the article file. A beat may be one sentence or several paragraphs — whatever that beat naturally is. Stop there.
 3. Re-read the article file from disk. Then offer 2–3 candidate **next beats** — different directions the journey could pivot to from where the article now stands.
 4. Loop steps 2–4 until the article reaches a natural end.
@@ -22,7 +22,7 @@ Then run a beat-by-beat journey:
 
 ## What is a beat
 
-A beat is one move in the journey. It does one thing — sets a scene, lands a point, asks a question, tells a small story, drops an aside, twists the angle. Then it stops, leaving the reader at a place where the next beat can pivot.
+A beat is one move in the journey. It does one thing — sets a scene, lands a point, asks a question, drops an aside, twists the angle. Then it stops, leaving the reader at a place where the next beat can pivot.
 
 A beat is sized by what it needs:
 
@@ -32,70 +32,21 @@ A beat is sized by what it needs:
 
 If a "beat" needs five paragraphs and three subheadings, it's not a beat — it's two beats glued together. Split it.
 
-## Offering candidate beats
-
-Each candidate should be genuinely different — different angle, different tone, different move. Not three flavours of the same paragraph. The user is choosing a _direction_, so the choices need to diverge.
-
-Format the offer like a menu:
-
-```
-Where do you want to start?
-
-1. **Open with the failure.** Drop the reader into the moment it broke —
-   the bug, the silence, the wrong number on the dashboard. Hooks on shock.
-
-2. **Open with the contradiction.** State the thing everyone believes,
-   then state the thing that turns out to be true. Hooks on curiosity.
-
-3. **Open with the small scene.** A specific morning, a specific
-   conversation. Hooks on intimacy.
-```
-
-Sketch the move, not the prose. The user picks a direction; you write the prose afterward.
-
-Always end the menu with your recommendation and a one-line reason. Don't sit on the fence — pick one. Example: "I'd go with **2** — the contradiction sets up the strongest through-line for what's in the pile." The user can override; they usually won't, but they need your read.
-
 ## Writing one beat
 
-Once a beat is picked, write _that beat only_ to the article file. Do not write the next beat. Do not foreshadow the next beat. Do not write transitions out of the beat — the next beat will pivot, and pivots are written when their beat is written.
+Once a beat is picked, write _that beat only_ to the article file. Do not write the next beat.
 
 Pull material from the raw pile to populate the beat. You can paraphrase, split, recombine, or quote. The pile is a quarry.
 
-If the beat needs something the pile doesn't have, name the gap before writing: "this beat wants a concrete example and the pile doesn't have one — give me one or pick a different beat."
-
-## Pivoting to the next beat
-
-After the user has edited, re-read the article file. The article may have changed in ways that change what the next beat should be. Then offer 2–3 candidates again.
-
-The candidates should respect the article so far. Useful pivot moves:
-
-- **Continue** — push further in the same direction, deepen what's there.
-- **Contrast** — introduce the opposite, the counterexample, the doubt.
-- **Zoom in** — narrow to a specific case, scene, or detail.
-- **Zoom out** — widen to the broader implication or pattern.
-- **Aside** — break the fourth wall, drop a tip, add a footnote-shaped paragraph.
-- **Pivot hard** — deliberately change subject, trusting the connection will land later.
-
-Mix the candidates. If you've offered three "continue" options in a row, force a contrast or zoom into the next menu.
-
 ## Ending the journey
 
 The article ends when the journey is complete — not when the pile is empty. Most piles will have leftover fragments that don't make it in. That is fine; that is the point of having more raw material than you need.
 
-When you sense an ending is near, say so: "we could end on the last beat, or add one more that lands the takeaway — which?" Let the user decide.
-
 ## Writing rhythm
 
 - Append one beat at a time. Never write ahead.
 - Re-read the article file from disk before every write. Preserve user edits absolutely.
-- If the user edits a previous beat substantially, let it change what comes next. The journey is alive.
+- If the user edits a previous beat substantially, let it change what comes next.
 - If the user says "rewrite that beat" or "go back and try a different beat 3", do it — edit in place, leave the rest alone.
 
-## Out of scope
-
-- Outlining the whole article up front.
-- Writing multiple beats in a single turn.
-- Editing the raw material file.
-- Imposing a fixed structure (intro/body/conclusion). The structure is whatever the journey turns out to be.
-
 </supporting-info>

skills/in-progress/writing-fragments/SKILL.md

@@ -11,6 +11,8 @@ As fragments emerge from either side of the conversation, append them to a singl
 
 If the user did not pass a path, ask once where to save the document, then remember it for the rest of the session.
 
+Capture fragments from the very first thing the user says, including the initial prompt.
+
 On first write, put a single H1 at the top with a working title (it can change later) and nothing else — no metadata, no TOC, no date.
 
 </what-to-do>
@@ -33,19 +35,6 @@ Fragments are deliberately heterogeneous. Examples of what could be a fragment:
 
 The novelist's diary is the model: years of unstructured noticings that later get mined for raw material. Fragments are noticings.
 
-## How the session feels
-
-Like a grilling session, with fragments as a by-product. The shape:
-
-- Start by asking what the user is circling around. Let them ramble.
-- Catch fragments from their words. When they say something quotable, capture it verbatim or slightly sharpened.
-- Generate candidate fragments back at them. Try things on. Offer two or three versions of the same idea and let them pick or reject.
-- When something lands, append it. When it doesn't, drop it.
-- Press for more. Excavate. "You said X — say it three different ways." "What's the version of that you wouldn't say in public?" "What's the example that made you believe this in the first place?"
-- Push on quality. The article will only be as good as its fragments. If a fragment is mushy, propose a sharper rewrite before appending.
-
-Blue-sky generously, but apply pressure to deepen and tighten. Do not rush to outline. Do not group. Do not propose a structure. If the user starts asking about structure, redirect: "We're still mining. Structure comes later."
-
 ## File format
 
 ```markdown
@@ -83,13 +72,4 @@ Before every write: re-read the file from disk. The user may have edited, reorde
 
 The user can say "cut the last one", "rewrite that one sharper", "merge those two" at any time. Treat those as first-class instructions.
 
-## Out of scope
-
-- Outlines, section plans, paragraph structure, transitions.
-- Deciding the article's thesis, argument, or audience.
-- Producing prose that reads as finished writing.
-- Linking fragments together with connective tissue.
-
-If the user is ready for any of that, the session is over and a different tool takes over. Do not name or recommend that tool — just stop.
-
 </supporting-info>