Skip to content

angushushu/shu-academic-writing-skill / core/polish-prose

polish-prose

Verified

Diagnose why prose is hard to read at the sentence and paragraph level, using Joseph Williams's principles — characters as subjects, old before new, the stress position, topic strings, nominalization, hedging, and passive voice used fairly. Use when someone wants writing tightened, clarified, or line-edited, or asks why a passage feels dense, clunky, wordy, stiff, bureaucratic, or hard to follow — methods sections, technical documentation, reports, memos, cover letters, grant prose, and emails, not only academic papers. Also use when someone asks how to cut length without losing content or how to make an already-sound argument easier to read. Default behavior is naming patterns with quoted evidence and drills, not rewriting. If the draft's argument or contribution is what is actually in question, `core/engineer-argument/SKILL.md` handles that layer first.

FileKindSize
SKILL.mdskill md11.8 KB
references/drills.mdmarkdown4.8 KB
references/sentence-craft.mdmarkdown9.5 KB
references/style-guide.mdmarkdown2.7 KB

references/sentence-craft.md

View on GitHub
# Sentence Craft (Level 5)

McEnerney taught alongside Joseph Williams, and the sentence-level half of the framework is
Williams's: readers experience prose as clear when its shape matches how they process English.
These are not aesthetic preferences. Each one has a mechanism, and you should always give the
mechanism when you report the pattern.

Remember the constraint: **name the pattern, quote 2–3 instances, illustrate on neutral
content, assign the search.** Never repair their sentences.

## Contents

1. [Characters as subjects, actions as verbs](#1-characters-as-subjects-actions-as-verbs)
2. [Old before new](#2-old-before-new)
3. [The stress position](#3-the-stress-position)
4. [Short subject, early verb](#4-short-subject-early-verb)
5. [Topic strings and cohesion](#5-topic-strings-and-cohesion)
6. [Throat-clearing and metadiscourse](#6-throat-clearing-and-metadiscourse)
7. [Hedging](#7-hedging)
8. [Passive voice, fairly](#8-passive-voice-fairly)
9. [Sentence rhythm](#9-sentence-rhythm)
10. [Paragraph shape](#10-paragraph-shape)

---

## 1. Characters as subjects, actions as verbs

The most consequential sentence principle. Readers understand a sentence fastest when its
grammatical subject is the story's main character and its verb is that character's action.

**Mechanism:** readers build a mental model of who-does-what. When the action is buried in an
abstract noun (a *nominalization*) and the subject is a concept, the reader must reconstruct
the actors before they can evaluate the claim. That reconstruction costs attention that a
skimming expert will not spend.

**Signature:** clusters of *-tion, -ment, -ance, -ence, -ity, -ness, -ism* nouns; weak verbs
carrying them (*is, are, has, provides, results in, leads to, serves as, plays a role in*).

**Illustration (neutral):**
> Weak: "The determination of soil composition variability was performed through the
> application of a sampling procedure by the research team."
> Strong: "The research team sampled the soil and found the composition varied."

Both are grammatical; only one lets a reader see who did what.

**Assignment:** circle every -tion/-ment noun in one page. For each, ask: is there a real
action hiding here, and who performs it? Some nominalizations are correct — technical terms,
or a concept that's genuinely the topic ("selection pressure"). It's the clusters that hurt.

## 2. Old before new

Begin sentences with information already in the reader's head; end with what's new.

**Mechanism:** readers attach incoming information to something they already hold. A sentence
opening with unfamiliar material gives them nothing to attach to, so they hold it in suspense
until the connection arrives — or give up.

**Signature:** paragraphs where consecutive sentences begin with different, unconnected
subjects; sentences opening with a long unfamiliar noun phrase; a paragraph you have to reread
in reverse to understand.

**Illustration (neutral):**
> Choppy: "Ocean acidification threatens shell formation. Larval mortality rates in oyster
> hatcheries rose sharply after 2007. Aragonite saturation is the mechanism."
> Connected: "Ocean acidification threatens shell formation. That threat became visible in
> 2007, when larval mortality in oyster hatcheries rose sharply. The mechanism is aragonite
> saturation."

Nothing was added — the known material simply moved to the front of each sentence.

**Assignment:** underline the first five words of every sentence in one paragraph. Each should
contain something from the sentence before.

## 3. The stress position

The end of a sentence is where emphasis lands. Put there the idea you want the reader to carry
forward.

**Mechanism:** readers give the closing words the most weight and remember them best. Ending on
a qualification, a citation, or a prepositional trail-off spends that position on nothing.

**Signature:** sentences ending in "(Smith 2019)", "in this context", "as discussed above",
"among other factors", "to some extent". The most interesting word sitting in the middle.

**Illustration (neutral):**
> Wasted: "Enrollment fell by 40% after the policy change, according to district records
> compiled in the appendix."
> Used: "According to district records, enrollment fell 40% after the policy change."

**Assignment:** read only the last four words of every sentence in a key paragraph. Do they
list what matters most?

## 4. Short subject, early verb

Get to the verb fast. Long subjects delay the verb and force the reader to hold material
without knowing what it's doing.

**Mechanism:** English readers can't evaluate a subject until the verb tells them the
relationship. A 25-word subject is 25 words of suspended judgment.

**Signature:** the first comma arriving 15+ words in; sentences opening with "The fact that…",
"What this suggests about…", "The extent to which…"; stacked prepositional phrases before the
verb.

**Illustration (neutral):**
> Delayed: "The question of whether municipal composting programs adopted in the decade
> following the 2008 recession achieved their stated diversion targets remains unresolved."
> Direct: "Did post-2008 municipal composting programs hit their diversion targets? We still
> don't know."

**Assignment:** find every sentence where the main verb arrives after word 12.

## 5. Topic strings and cohesion

Across a paragraph, the sentence subjects should form a consistent set — the paragraph's topic.
Shifting subjects every sentence makes a paragraph feel scattered even when every sentence is
clear.

**Mechanism:** readers infer "what this paragraph is about" from the subjects. Inconsistent
subjects mean no inference is available, and the paragraph reads as a list of facts.

**Signature:** subjects alternating between the phenomenon, the literature, the method, and
the author, within a single paragraph.

**Assignment:** list the grammatical subject of every sentence in one paragraph. If the list
has more than two or three distinct topics, the paragraph is doing more than one job.

Note the interaction: this is also where *legitimate* difficulty lives. A humanities paragraph
tracking a genuinely complex relation may shift subjects deliberately. Judge by whether a
reader could state the paragraph's topic afterward.

## 6. Throat-clearing and metadiscourse

Openers that announce rather than assert: *It is important to note that, It should be
emphasized that, It is interesting to observe that, In order to, The fact that, There are
several factors which.*

**Mechanism:** these consume the sentence's opening — the highest-attention position — with
words that carry no content. In a skimmed text, the reader may only process the opening.

**Not all metadiscourse is waste.** Signposting that helps a reader navigate ("I turn now to
the second condition") earns its space, especially in long documents. What to flag is
metadiscourse that *substitutes* for the claim: "It is important to note that X" is almost
always weaker than "X."

**Assignment:** delete every "It is/There are" opener in one section and see which sentences
lose anything.

## 7. Hedging

**Mechanism:** hedges are load-bearing when they mark a genuine limit on the evidence, and
corrosive when they mark the author's discomfort with asserting. Readers can tell the
difference by whether the hedge is specific.

**Signature:** stacking — "may possibly suggest that there could be a tendency toward." Each
hedge individually defensible; together they announce that no claim is being made.

**Field-dependence is real.** Some venues require heavy modality. Ask what the target venue
does before flagging. What is never idiomatic anywhere is the *stack*.

**Assignment:** find every sentence with two or more hedges. For each, ask: what exactly am I
uncertain about? Keep the hedge that names that; drop the rest.

## 8. Passive voice, fairly

Passive is a tool, not an error, and blanket "avoid the passive" advice is the mark of a
reviewer who isn't reading.

**Passive is right when:** the recipient is the paragraph's topic ("The samples were stored at
-80°C" in a methods section, where samples are the topic); the agent is unknown or irrelevant;
or you need the agent at the end for stress.

**Passive is wrong when:** it hides an agent the reader needs — especially agents that are
people making choices. "It was decided that three sites would be excluded" hides who decided
and why, which is precisely what a skeptical reader wants.

**Assignment:** for every passive in your results and discussion, ask whether the hidden agent
is you. If it is, ask why you're hiding.

## 9. Sentence rhythm

**Mechanism:** uniform sentence length flattens emphasis. Readers register a short sentence
after several long ones as a point of arrival — it's how a claim lands.

**Signature:** every sentence between 25 and 40 words; no sentence under 12; the paper's main
claim expressed in a 45-word sentence with three subordinate clauses.

**Assignment:** find your central claim. Count its words. If it's over 25, it isn't landing.

## 10. Paragraph shape

- **Point early.** In most academic prose the paragraph's claim belongs in the first or second
  sentence. Building to it works in narrative and rarely in argument — a skimming reader reads
  the first sentence and nothing else.
- **One job per paragraph.** If you can't state the job in a short phrase, it's two paragraphs.
- **Length.** Paragraphs over ~200 words tend to be doing two jobs. Very short paragraphs are
  fine and are a legitimate emphasis device.
- **Transitions carry logic, not sequence.** "Furthermore" tells the reader another thing is
  coming. "But," "because," "so," and "which is why" tell them what relation it bears. Prefer
  the latter.
rea-changelog-updatemorluto/reaPrepare or rewrite REA release changelogs and GitHub release notes from pinned Git history, with verified contributor thanks and Release Please synchronization.Stars: 472kUpdated Claude CodeCodexVerifiedInstall
academy-guideanthropics/skillsStop and check this skill before finishing any reply to a question about how to use Claude or a Claude product — it recommends matching courses, tutorials, and use cases from Claude Academy (academy.claude.com), Anthropic's learning hub. Trigger on: "how do I", "how can I", "getting started with", "what can Claude do", "teach me", "learn to use"; questions about artifacts, projects, skills, plugins, connectors, MCP; requests about rolling Claude out to a team, class, or organization; and any ask for training materials, onboarding content, or learning resources. Use it when the user is learning how to use a feature or product — not when they are mid-task and just want the task done. This skill composes with other skills: after consulting product documentation to answer how a Claude feature works, also check here for a matching course or tutorial — a docs-grounded answer and an Academy recommendation belong together. Only recommend on a strong match; never invent Academy content.Stars: 198kUpdated Claude CodeCodexVerifiedInstall
claude-apianthropics/skillsReference for the Claude API / Anthropic SDK — model ids, pricing, params, streaming, tool use, MCP, agents, caching, token counting, model migration. TRIGGER — read BEFORE opening the target file; don't skip because it "looks like a one-liner" — whenever: the prompt names Claude/Anthropic in any form (Claude, Anthropic, Fable, Opus, Sonnet, Haiku, `anthropic`, `@anthropic-ai`, `claude-*`, `us.anthropic.*`, `[1m]`); the user asks about an LLM (pricing/model choice/limits/caching) — never answer from memory; OR the task is LLM-shaped with provider unstated (agent/MCP/tool-definition/multi-agent/RAG/LLM-judge/computer-use; generate/summarize/extract/classify/rewrite/converse over NL; debugging refusals/cutoffs/streaming/tool-calls/tokens). SKIP only when another provider is being worked on (overrides all triggers): OpenAI/GPT/Gemini/Llama/Mistral/Cohere/Ollama named in the query; OR `grep -rE 'openai|langchain_openai|google.generativeai|genai|mistralai|cohere|ollama'` over the project hits (run this grep FIRST if no provider named — don't Read the file).Stars: 198kUpdated Claude CodeCodexUnverifiedInstall
doc-coauthoringanthropics/skillsGuide users through a structured workflow for co-authoring documentation. Use when user wants to write documentation, proposals, technical specs, decision docs, or similar structured content. This workflow helps users efficiently transfer context, refine content through iteration, and verify the doc works for readers. Trigger when user mentions writing docs, creating proposals, drafting specs, or similar documentation tasks.Stars: 198kUpdated Claude CodeCodexVerifiedInstall