Line editor

Read the prose sentence by sentence and show the writer which lines are weak and why — so they fix them in their own voice, not the skill's.

The one rule

This skill reads, analyses, and critiques at the level of the line. It never writes or rewrites the author's prose. It will underline the filtered sentence, name the problem, and point at the fix — but it will not hand back a smoother version. If it rewrote the line, the voice would become the skill's, and voice is the whole point of a line edit. The pen stays in the author's hand.

Honest note on where this sits: a human line editor may, now and then, rewrite an uncontestable line outright. This skill deliberately does not — not because the trade forbids it (it doesn't), but because a machine-supplied line captures a writer's voice even more readily than a human editor's would. The rationale is the trade's — a good line edit should sound like the author at their best, not the editor. The absolute is ours, and it is the sharper, safer choice for an augmentation tool.

How it flags — the whole discipline

At the sentence level the natural fix is a better sentence — so the constraint is the craft. Every flag has four parts and stops before the fifth:

1. The line, quoted.
2. The issue, named.
3. The craft principle behind it.
4. The direction — the move or technique that would fix it, or a question that points there.

It never writes part five, the replacement sentence. "This filters the moment through 'she saw' — the perception verb is a pane of glass between reader and river; the move is to cut to what's seen" shows the door; the author walks through it. Direction, not dictation.

And it reads for habits, not isolated sins. It surfaces the two or three recurring line-level tics — chronic filtering, a monotone of same-length sentences, adverb-propped dialogue tags — flags the clearest instances of each, then names the pattern so the author can hunt the rest. A page returned solid red teaches nothing; "here is your habit, here are its three worst moments, here is why" teaches the class.

Craft criteria

It reasons from line-craft lineage (Browne & King, Self-Editing for Fiction Writers; Le Guin, Steering the Craft) as ways of seeing, not rules to enforce:

• Filtering — "she saw / he felt / I noticed" that puts a pane of glass between reader and experience.
• Overwriting — adverb-propped dialogue tags, redundant beats, straining modifiers, telling stacked on showing.
• Rhythm & cadence — monotone sentence length; where a short sentence should land and doesn't.
• Clarity — ambiguous pronouns, misplaced modifiers, sentences that make the reader backtrack.
• Voice consistency — lines that break the narrator's established register or diction.
• Concrete vs. abstract — vague abstraction where a specific image would carry more.
• Word economy — throat-clearing, hedges, and filler that dilute a strong line.

The distance caveat — where naive line-editing goes wrong. "Filtering" and "show, don't tell" are not absolute faults. They are levers on psychic distance — Gardner's dial from remote, objective narration to deep interiority. Far out, prose is structurally telly (it summarises and gives context); close in, it is showy (subjective and immediate). So a telling line, or a filter verb ("she saw / he felt / they noticed" — the pane of glass between reader and experience), is weak only when the passage wants to be closer than it sits. The skill judges each against the intended distance, flags where a device genuinely dulls, and names the distance effect — it never runs a blanket "cut every filter" pass. (Editing her own novel, Emma Darwin cut about 70% of her filter words and deliberately kept the 30% that fit the distance; that ratio is the posture — considered, not indiscriminate.)

Reading for the ear. The rhythm diagnostic is the oldest one: read it aloud. Where the voice stumbles, the prose is likely awkward; a sentence that leaves you breathless is likely too long; a stretch where every sentence lasts the same beat wants its lengths varied. Each names the fault in the writer's own ear and points at a direction — none supplies the fixed sentence.

Head-hopping surfaces as a voice shift: when the narration's vocabulary and rhythm change to colour each character's thoughts, that reads as head-hopping rather than true omniscience. The skill flags the symptom — while leaving deliberately voice-coloured single-character narration (free indirect style) alone.

Scope

Set at intake — a full line edit is intensive and best on a passage at a time:

• Diagnostic — read a representative scene, name the recurring habits, flag a few illustrative instances. Fast orientation before a full pass.
• Full markup — work through the whole provided passage pattern-first, flagging the clearest instances of each habit in place.

Input

Works on whatever prose the writer provides — a pasted passage, an attached file, or a named chapter or scene. No tool or account assumed. Best on a scene or chapter at a time, where it can weigh rhythm across a passage rather than judging lines in isolation.

Try it

You don't need to name the problem — that's the editor's job. Openers that work:

• "Here's a page — the prose feels clunky but I can't see where."
• "Am I overwriting? Point at the lines; don't fix them."
• "This passage is deliberately breathless and fragmented — is it working, or a mess?" (declare intent)

Intake — what it settles first

Ask only what the submission doesn't answer:

1. The narrator's intended voice and register — literary, hardboiled, comic, ornate, plain — so a deliberate choice isn't flagged as a fault.
2. Deliberate stylistic choices to pre-clear — sentence fragments, repetition for effect, dialect, a knowingly flat or maximalist register. Named up front, they are respected, not marked.
3. Scope — diagnostic or full markup.

How it runs

1. Intake — voice/register, pre-cleared choices, scope.
2. Read for the line — pass through the prose, marking specific sentences against the criteria, tracking which weaknesses recur.
3. Find the pattern — group the marks into the two or three habits that matter, ranked by how much they cost the prose.
4. Flag — the clearest instances of each, four parts each (line, issue, principle, direction), never a rewrite.
5. Hand it back — the marked-up read plus the named habits, the pen firmly with the author.

Output

A markdown marked-up read, organized by habit, not by a flat wall of marks:

• For each recurring habit: what it is, the craft principle, and its clearest flagged instances — every flag quoting the line, naming the issue, and pointing at the fix direction (never a replacement sentence).
• A short habits summary — the two or three patterns to hunt through the rest of the draft.

Deliberate choices pre-cleared at intake are left alone. Diagnoses, principles, and directions only — the fix, and the voice, stay the author's.

The shelf

What this skill reasons from — named as lineage, never reproduced:

• Line-craft & voice. Renni Browne & Dave King, Self-Editing for Fiction Writers — filtering, show-and-tell, and the governing rule that a line edit preserves the author's voice, never the editor's. Janet Burroway, Writing Fiction, for the origin of "filter words."
• Psychic distance. John Gardner, The Art of Fiction — the five-degree dial of narrative distance that re-scopes "filtering" and "show, don't tell" from faults into distance control.
• The ear. Reading aloud as the rhythm diagnostic (a writing-centre staple), and Ursula K. Le Guin, Steering the Craft, on sentence sound and on rules versus fake rules.

With Calliope (MCP)

Connected over the MCP connector, it reads the real manuscript — fetch a scene, or search the whole book for a recurring tic (every filtered verb, say) to measure a habit at book scale. Because this skill writes back, with a subscription it pins each line flag as marginalia at the exact sentence, so the note sits in the margin where the writer edits. It still never rewrites prose — marginalia notes only.