Skip to content

Output style and accessibility

Nyia tells every assistant to answer in a fixed, short, explicit style. It suits all developers, and it is built for ADHD, autism, giftedness and dys profiles. There is no profile to pick, so you never disclose a diagnosis.

What you get

  • The answer comes first, in 1-2 sentences.
  • Sentences have 25 words at most, and paragraphs 3 sentences at most.
  • Facts are restated, never replaced by "as above". Acronyms are defined at first use.
  • There is no filler, no praise and no narration.
  • Task answers, summaries and idea discussions use one block:
Status: fixed, 12 of 12 tests pass.
What: the login crash is gone (`src/auth.js:45`).
Why: the API returns null for deleted users.
Next: deploy `src/auth.js`.
  • A one-line fact is only the answer. A question to you comes with why it matters.
  • Ambiguous requests get an answer plus a line such as Assumption: you mean the staging server.
  • After a short answer you may see Go deeper on: the cache, the retry logic?
  • A deep dive ends with Back to the main task: and 3 bullets at most: goal, status, next.

Labels are written in the language of the answer.

Tables

Recurring data always uses the same headers, with one line of context above the table.

Data Headers
Plans or work Item / State / Next
Findings Severity / File:line / Problem / Fix
Options Option / Gain / Cost / Risk
Test results Suite / Pass / Fail / Note

Emojis and humour

Unicode emojis are off by default. Turn them on with output_emoji:

nyia config global output_emoji=on      # every project
nyia config project output_emoji=off    # this project; wins over global

An invalid value gives a warning and off. Humour stays on: it is a short aside, marked with an ASCII emoticon such as ;).

When the style does not apply

Case What happens
Files, commits, code, tool input The style never applies
Exact format required (JSON, a template, a parser contract) The format wins
A skill, persona or team template Its own format wins for that output
You ask in the chat ("long version", "reply in JSON") Your request wins for that answer
Automated review prompts They use their own fixed prompts, without this style

Prompt override files cannot remove the style. It is Nyia's base behaviour, not a security rule.

Limits

  • The rules guide a model. They do not guarantee every answer.
  • Vibe: the composed prompt includes the style, but it does not reach the Vibe model yet.