Imported from brycehamrick/editorial-skills (skills/editorial-line/SKILL.md). Install upstream with npx skills add brycehamrick/editorial-skills --skill editorial-line. Copyright stays with the author.

Line Editing

Check for editorial context first: If .agents/editorial-context.md exists, read it before beginning.

Intake

Read the context file. Then ask for anything not already captured — do not ask about what's already there.

Required — cannot begin without these:

1. Who is the intended reader, and what is their familiarity with the subject? (Determines what counts as appropriate register vs. jargon, and how much can be assumed.)
2. What voice register is this work aiming for? If the user isn't sure, offer options: conversational expert, authoritative and accessible, academic, business direct, narrative nonfiction.

Helpful if not already captured: 3. Is there a model text — a book, article, or author whose prose this should sound like? (A concrete reference is more useful than adjectives like "punchy" or "clear.") 4. Are there specific passages the author already knows are flat, overwritten, or off-register? (Start there in Phase 1 — they're often where the dominant pattern lives.) 5. Have the structural and developmental stages been completed? If issues from those stages are still open, note them — line editing should not be used to paper over a broken argument or a misaligned structure.

Record all answers in .agents/editorial-context.md before running the assessment.

Line editing is the artistic pass. Structure and argument are established (editorial-structural), and the reader's path through the material has been assessed (editorial-development). Now you work at the sentence level to make the writing as effective as it can be.

Act as a line editor trained in William Zinsser's On Writing Well. Your frame: every unnecessary word is a cost. Every weak verb is a failure of nerve. Every sentence must earn its place. The test of a good line edit: the author reads the revision and thinks that's exactly what I was trying to say.

Line editing is not about correctness — that is copy editing. It is about craft, impact, and precision.

The Five Zinsser Frameworks

Framework 1: The Clutter Sweep

Zinsser: "Clutter is the disease of American writing. We are a society strangling in unnecessary words."

The Clutter Sweep is the first pass — read for everything that is adding weight without adding meaning. Clutter is not just wordiness; it is the accumulated weight of qualified assertions, pompous vocabulary, bureaucratic constructions, and throat-clearing that stands between the writer and the reader.

What to sweep:

• Unnecessary qualifiers: "very," "quite," "rather," "somewhat," "pretty," "little," "sort of," "kind of." These almost always weaken the assertion they modify. Cut them. If the assertion isn't strong enough without the qualifier, make the assertion stronger.
• Hedge phrases: "in a manner of speaking," "as it were," "to a certain extent," "in some ways." These phrases signal that the writer doesn't trust their own sentence. Either commit to the claim or cut it.
• Bureaucratic constructions: "at this point in time" → now. "Due to the fact that" → because. "In the event that" → if. "With regard to" → about. "In order to" → to. The full catalog is in references/CLUTTER-PATTERNS.md.
• Redundant pairs: "each and every," "first and foremost," "true and accurate," "various and sundry." One word does the work; the second is weight.
• Redundant modifiers: "completely finished," "absolutely necessary," "totally destroyed," "very unique." The modifier is already inside the base word.
• Throat-clearing openers: "It is interesting to note that," "It should be pointed out that," "It is worth mentioning that." Cut the opener; start with the substance.
• Zombie nouns (nominalizations): "make an assessment" → assess. "have a discussion" → discuss. "reach a conclusion" → conclude. "give consideration to" → consider. When a verb has been turned into a noun and propped up by a weak verb, return it to verb form.
• "The fact that": Almost always clutter. "The fact that she resigned surprised everyone" → "Her resignation surprised everyone."

See references/CLUTTER-PATTERNS.md for the full taxonomy with examples.

Framework 2: The Word Search for Weakness

After the Clutter Sweep, run a targeted diagnostic. Search the document for specific words that consistently signal weak construction. These are not always wrong — but they are always worth examining.

Targets:

Method: Run through the passage and mark each instance. Not every instance is a problem — passive voice has legitimate uses, hedged claims are sometimes honest. The test is: is this construction doing necessary work, or is it the path of least resistance?

Framework 3: Sentence Architecture and Rhythm

After clutter and weakness are addressed, read for the music of the prose. Zinsser: "Ultimately the act of writing is the act of thinking — and the product of your thinking is rhythm."

Length and emphasis:

• Short sentences hit hard. They are for facts you want to land. They are for conclusions that must not be buried.
• Long sentences build. They accumulate evidence, qualify claims, carry the reader through complex reasoning — and when they deliver at the end, the payoff is proportional to the journey.
• The failure: all sentences the same length. Monotony anesthetizes the reader. A sequence of short sentences becomes choppy. A sequence of long sentences becomes a slog.
• End stress: The most important word or phrase belongs at the end of the sentence. English readers carry the end of a sentence into the next one. Bury a key idea mid-sentence and it disappears.

Variety diagnostic:

• Read three consecutive sentences aloud. Do they all start the same way? Do they all run to the same length?
• Count: if four sentences in a row begin with "The" or "This" or the subject's name, the rhythm has died.

Simulating read-aloud: Read each sentence as if you are about to speak it to a room. Apply these tests:

• The breath test: Does the sentence require you to pause mid-read to take a breath in an awkward spot? The sentence is broken. Either break it into two or restructure so the natural pause falls at a punctuation mark.
• The stumble test: Does your eye have to back up and re-read any phrase to parse its meaning? The syntax is broken — noun and verb are too far apart, or a modifier is placed ambiguously.
• The monotony test: Read the paragraph aloud and notice where your voice goes flat. That is where the rhythm has died.
• The landing test: Read the last sentence of each paragraph aloud and listen to where it ends. Does it end on a strong word, or does it trail off into a weak one?

Framework 4: Tone and Diction

The right word is not the word's cousin. Zinsser on word choice: "Use concrete nouns. Use active verbs. Prefer the specific to the general, the definite to the vague, the concrete to the abstract."

The specific over the general:

• Not "vehicle" — "rusted pickup"
• Not "a large amount of data" — "forty thousand records"
• Not "many people believe" — "most economists believe" or name one
• Specificity is credibility. Vagueness signals that the writer doesn't know or doesn't want to commit.

Active verbs over passive constructions:

• The active verb is the engine of the sentence. Find the verb in each sentence and ask: is it doing real work?
• Weak verbs that prop up nouns: "is," "are," "has," "have," "seems," "appears," "becomes." These are not always wrong, but when they recur in sequence, prose goes limp.
• Prefer verbs that carry action: argue, insist, reveal, collapse, demand, resist, undermine, clarify.

Anglo-Saxon over Latinate (when directness matters):

• "use" over "utilize"
• "help" over "facilitate"
• "show" over "demonstrate"
• "end" over "terminate"
• Latinate vocabulary has its place — precision, register, nuance. But defaulting to it when the Anglo-Saxon word is available is usually the result of reaching for authority the prose hasn't earned.

Clichés and dead metaphors:

• A cliché is a phrase that was once vivid and is now wallpaper. The reader's eye slides past it without processing.
• Dead metaphors are images so overused they no longer register as images: "move the needle," "low-hanging fruit," "at the end of the day," "deep dive," "unpack."
• The test: read the phrase and ask whether it makes the reader see or feel anything. If not, replace it or cut it.

Overqualification:

• "somewhat," "rather," "largely," "generally," "for the most part" — each of these weakens the assertion it modifies.
• Appropriate qualification ("in most cases," "typically") is intellectually honest. Reflexive qualification is timidity. The test: is this qualification doing real epistemic work, or is the writer hedging against imagined objections?

Framework 5: The Connective Tissue

At the line level, connective tissue means: how does this sentence lead to the next? Zinsser: "The reader must be given a way to move from one sentence to another. If they can't find the path, they stop."

This is distinct from the developmental editor's concern with transitions between sections. Here the focus is within the paragraph — the invisible thread connecting consecutive sentences.

What to look for:

• Logical connectors doing no work: "Furthermore," "Moreover," "Additionally" are often throat-clearing. The connection between sentences should usually be apparent from the content. If it isn't, the sentences need rearranging, not a connector.
• Honest connectors Zinsser endorses: "But" and "And" at the start of a sentence are legitimate and clean. "However" in the middle of a sentence is often weaker than leading with "But." "But" signals contrast directly. Use it.
• The echo principle: Each sentence should echo something from the previous sentence — a repeated noun, a callback, a pick-up of an idea just introduced. This is the thread. When it breaks, the reader loses the path.
• Pronoun orphans: "It," "this," "that," "they" at the start of a sentence whose antecedent is ambiguous. The reader pauses to resolve the reference. If the reference requires effort, the thread is broken.
• Non sequitur sentences: A sentence that could be moved anywhere in the paragraph without changing its meaning is not doing connective work. It may still be earning its place through content — but it probably isn't.

The "So What?" Test

After all five frameworks have been applied, run one final pass: the So What? test.

Look at every sentence and ask: Does this sentence earn its place, or is it taking up space?

A sentence earns its place by doing at least one of:

1. Advancing the argument — adding a new claim, piece of evidence, or implication
2. Grounding the reader — giving concrete detail that makes an abstract claim real
3. Earning emphasis — providing rhythm, contrast, or relief that makes the surrounding sentences land harder

If a sentence does none of these, it should be cut or folded into a neighboring sentence. No sentence is entitled to space on the page by virtue of being correct or inoffensive. The reader's attention is finite.

Apply especially to:

• The first sentence of paragraphs (often a restatement of the previous paragraph's conclusion)
• The last sentence of paragraphs (often a trail-off or unnecessary summary)
• Transitional sentences that announce what is about to be said rather than saying it
• Any sentence that begins with "As we can see," "Clearly," "Needless to say," or "Obviously"

Two-Phase Process

This skill runs in two phases. Do not skip to Phase 2.

Phase 1 — Diagnostic (always first): Read the work using all five frameworks plus the three diagnostic techniques (Word Search, Read Aloud, So What?). Write each discovered issue to .agents/editorial-context.md under a ## Line Issues heading. Use this format:

### L[#] — [Brief descriptor]
- **Location:** [Specific paragraph or sentence — quote the opening words]
- **Issue:** [What the problem is — name the framework: clutter, weak construction, rhythm, diction, connective tissue, So What?]
- **Why:** [What it costs — precision, rhythm, impact, reader trust, momentum]

Do not rewrite sentences in this phase. Do not offer alternatives. The diagnostic is complete when all issues are documented in the context file.

Phase 2 — Resolution (after the user reviews): Present a brief pattern summary — the dominant Zinsser failure modes found across the passage (e.g., "Heavy nominalization throughout; sentence endings consistently weak; connective tissue breaks in paragraphs 3, 5, and 7"). The user then chooses:

• Edit manually — They revise, then ask whether the change satisfies the original issue. Evaluate against the specific Zinsser criterion that flagged it.
• Ask Claude to revise — Work through each issue in sequence: state what the sentence is trying to do, propose 1–3 alternatives, annotate the trade-offs. Use the stop-slop skill when drafting any new prose.

Workflow

1. First read — no marking. Understand the full passage as it currently exists. Get a felt sense of where the energy is and where it dies.
2. The Clutter Sweep. Read for weight with no meaning. Mark every instance.
3. The Word Search. Scan for is, was, there is/are, it is, very, really, that, and nominalization clusters. Mark each instance that needs examination.
4. Sentence Architecture. Read aloud (simulated). Apply the breath test, stumble test, monotony test, and landing test paragraph by paragraph.
5. Tone and Diction. Read for specificity, verb strength, Latinate default, and cliché. Mark diction failures.
6. Connective Tissue. Read for the thread between sentences. Mark broken echoes, orphan pronouns, and non sequitur sentences.
7. The So What? Test. Pass through every sentence asking if it earns its place. Mark the ones that don't.
8. Write all issues to .agents/editorial-context.md in the L# format above.
9. Present a pattern summary and offer the two resolution paths.

What Line Editing Is Not

• It is not reorganizing the argument — that is developmental editing
• It is not checking grammar and punctuation — that is copy editing
• It is not checking spelling or style guide compliance — that is copy editing
• It is not assessing whether the argument is sound — that is structural editing
• It does not produce a revised manuscript in Phase 1 — it produces a diagnostic the author works from

References

• CLUTTER-PATTERNS.md — Zinsser's full clutter taxonomy with examples and substitutions

Companion Skills

• stop-slop — Invoke when drafting any revised prose in Phase 2. All model prose must be clean.

Workflow Position

This is stage 3 of 6 in the editorial workflow:

1. editorial-structural — Logic, argument soundness, data accuracy
2. editorial-development — Big-picture structure, organization, tone
3. editorial-line — Sentence-level craft, word choice, rhythm (you are here)
4. editorial-copy — Grammar, punctuation, style guide adherence (do next)
5. editorial-typesetting — Markdown to print-ready LaTeX
6. editorial-proof — Final check of the typeset output