Profile Writing Style
Use this skill to infer a writer's recurring habits from text they provide and
turn those observations into a concise Markdown style profile. Focus on
describing actual tendencies, not on scoring quality or rewriting the samples
unless the user asks for that separately.
Workflow
1. Build the sample set
- Prefer 3 to 10 files or excerpts written by the same person.
- If the user gives a directory, identify files that contain the writer's own
prose and ignore generated files, bibliographies, logs, figures, and code.
- If the sample set mixes genres, state that clearly and group observations by
genre when needed.
2. Read for recurring patterns
Look for patterns that repeat across the samples rather than isolated quirks.
Summarize only what is supported by multiple examples.
Check these layers:
- sentence shape: average sentence length, clause density, use of contrast or
qualification, rhythm variation
- paragraph movement: topic sentence style, buildup, paragraph endings,
tendency to summarise or restate
- argument style: directness, hedging, evidence handling, how claims are
limited or framed
- vocabulary: concrete versus abstract nouns, repeated framing words,
transition words, domain vocabulary
- meta-writing: how often the writer refers to the paper, report, or section
itself instead of advancing the object-level claim
- tone: restrained, assertive, explanatory, defensive, formal, compressed, or
expansive
3. Distinguish stable style from local noise
- Do not overfit to one chapter, one assignment, or one unusually edited file.
- Separate genuine style from genre constraints. For example, notation-heavy
technical sections naturally sound more templated than discussion sections.
- If the samples appear heavily AI-assisted or heavily edited, say so as an
uncertainty instead of pretending the profile is pure.
4. Produce a Markdown style profile
Write a Markdown document using the structure in
references/style-profile-template.md.
The profile should include:
- a short overview of the writer's overall prose character
- 5 to 10 concrete stylistic tendencies
- a short list of high-frequency habits or repeated constructions
- tensions or tradeoffs in the style, such as "clear but repetitive" or
"careful but over-explained"
- a short "Strengths To Preserve" section
- a short "Risks To Watch" section
- optional "Operational Rules" if the user wants the profile reused later
5. Keep the output evidence-based
- Ground claims in patterns actually seen across the sample set.
- Use short quoted fragments only when helpful and keep them brief.
- Avoid vague labels like "good academic tone" unless you explain what creates
that effect in the writing.
- Do not turn the profile into a rewrite request unless the user asks.
Output Notes
- Default output file name:
writing-style-profile.md unless the user names a
different path.
- If the sample size is too small, say what can and cannot be inferred.
- If the user wants a reusable prompt or editing guide derived from the
profile, add a final
Operational Rules section.
1---2name: profile-writing-style3description: Use when the user wants an AI to analyze one or more text files written by them and summarize their writing style into a Markdown document. Trigger on requests like "summarize my writing style", "analyze these samples and describe how I write", "extract my prose habits from these files", or "create a style profile from my writing".4---56# Profile Writing Style78Use this skill to infer a writer's recurring habits from text they provide and9turn those observations into a concise Markdown style profile. Focus on10describing actual tendencies, not on scoring quality or rewriting the samples11unless the user asks for that separately.1213## Workflow1415### 1. Build the sample set1617- Prefer 3 to 10 files or excerpts written by the same person.18- If the user gives a directory, identify files that contain the writer's own19 prose and ignore generated files, bibliographies, logs, figures, and code.20- If the sample set mixes genres, state that clearly and group observations by21 genre when needed.2223### 2. Read for recurring patterns2425Look for patterns that repeat across the samples rather than isolated quirks.26Summarize only what is supported by multiple examples.2728Check these layers:2930- sentence shape: average sentence length, clause density, use of contrast or31 qualification, rhythm variation32- paragraph movement: topic sentence style, buildup, paragraph endings,33 tendency to summarise or restate34- argument style: directness, hedging, evidence handling, how claims are35 limited or framed36- vocabulary: concrete versus abstract nouns, repeated framing words,37 transition words, domain vocabulary38- meta-writing: how often the writer refers to the paper, report, or section39 itself instead of advancing the object-level claim40- tone: restrained, assertive, explanatory, defensive, formal, compressed, or41 expansive4243### 3. Distinguish stable style from local noise4445- Do not overfit to one chapter, one assignment, or one unusually edited file.46- Separate genuine style from genre constraints. For example, notation-heavy47 technical sections naturally sound more templated than discussion sections.48- If the samples appear heavily AI-assisted or heavily edited, say so as an49 uncertainty instead of pretending the profile is pure.5051### 4. Produce a Markdown style profile5253Write a Markdown document using the structure in54[references/style-profile-template.md](references/style-profile-template.md).5556The profile should include:5758- a short overview of the writer's overall prose character59- 5 to 10 concrete stylistic tendencies60- a short list of high-frequency habits or repeated constructions61- tensions or tradeoffs in the style, such as "clear but repetitive" or62 "careful but over-explained"63- a short "Strengths To Preserve" section64- a short "Risks To Watch" section65- optional "Operational Rules" if the user wants the profile reused later6667### 5. Keep the output evidence-based6869- Ground claims in patterns actually seen across the sample set.70- Use short quoted fragments only when helpful and keep them brief.71- Avoid vague labels like "good academic tone" unless you explain what creates72 that effect in the writing.73- Do not turn the profile into a rewrite request unless the user asks.7475## Output Notes7677- Default output file name: `writing-style-profile.md` unless the user names a78 different path.79- If the sample size is too small, say what can and cannot be inferred.80- If the user wants a reusable prompt or editing guide derived from the81 profile, add a final `Operational Rules` section.