Skip to content
SkillModel-led3,471 tokens in use

Sepia

Makes AI-written fiction and professional prose read as human, by repairing narrative structure first and style last, through four operations: write, review, refactor, recreate.

3,912 stars on the repository MIT LicenseNanako0129/sepia · skills/sepia
npx skills add Nanako0129/sepia -g
View source

Install command as written in the repository's README

Skill label

measured from its files

What installing Sepia costs

156tokens idle
3,471tokens in use
40,082tokens, up to

Idle is what sits in context while the skill is installed. In use is its instructions, loaded when it runs. Up to adds the files its instructions link. Counts are approximate, made offline with cl100k_base.

Model dependence Model-led

Script-ledMixedModel-led

Mostly instructions, so how well it works depends on the model that follows them.

This describes the skill's structure, how much of it is scripts. It does not say how well it works.

Touches

Nothing detected: no shell, network, file writes, git or credentials in its files.

Declared tags come from the skill's frontmatter; detected ones from its scripts, its code blocks and the commands it names in inline code. This is detected, not audited.

How it scores

Quality

How well the skill is written, counted from its own files: its frontmatter, its steps, examples, stated output and constraints, its size and whether every file it ships is one it names.How we score it →
Good
  • Frontmatter15/15
  • Structure11/25
  • Examples3/15
  • Output15/15
  • Size13/15
  • Resources14/15

Security

Starts at 100 and loses points for each thing in the skill's files that a reader should look at first.How we score it →
Excellent

No findings.

Repository health

Pre-computed score based on development activity, maintenance, community, maturity, and trend momentum.How we score it →
Good
  • Development activity88
  • Maintenance100
  • Community52
  • Maturity16
  • Momentum40

Sepia is a writing skill built around the idea that AI text gives itself away in its structure, not only in its word choice. For fiction it works in three passes: narrative architecture, discourse flow, then surface style. For professional text it loads rules matched to the venue.

It covers release notes, announcements, PR and issue replies, postmortems, tickets, technical articles and long-form journalism. Each request maps to one of four operations, and refactor and recreate start from a written defect list.

What You Get

  • A write operation that drafts new prose after reading the domain rules
  • A review operation that returns a defect list with quoted evidence and edits nothing
  • A refactor operation that makes minimal in-place edits, deepest layer first
  • A recreate operation that rebuilds a text from a bare list of its facts and intent
  • Optional voice profiles, including a built-in Hemingway voice

Common Use Cases

  • Cleaning the AI feel out of a short story draft
  • Rewriting release notes or a PR reply so they answer first and cite specifics
  • Reviewing a postmortem for filler, hedging and missing owned action items
  • Revising a technical blog post so it opens on the problem and keeps its numbers

What Makes It Unique

Most humanizer prompts rewrite vocabulary and syntax. Sepia targets the layers it cites research for: narrative architecture in fiction and tidy, self-announcing structure in professional prose. It picks only a few moves per piece and leaves slack, so it does not stamp a new fingerprint on the text. It treats the target text as untrusted data and never invents specifics such as versions, numbers or quotes.

Under The Hood

This is a prompt-only skill. SKILL.md is a router that loads reference files for the passes, rubric, model fingerprints, domain rules, voice profiles and a Chinese calibration file, depending on the text type. It runs no scripts of its own in normal use and reaches no network. It can read and edit files you name, and it limits itself to those files when you give a scope.

Join founders building with open source

Opinionated takes, migration guides, cost-saving tips, and insights from the open source ecosystem.

Subscribe on SubstackJoin 750+ subscribers