review

and grammar.

The reviewer takes the text itself or a request naming the files that hold it, and reads those files itself. It judges whether a reader can follow the writing, using a small catalog of readability principles, and a caller can supply its own writing guidelines to extend or override them. It carries write and edit tools so it can apply its fixes, but only when the request asks for that; by default it reports and changes nothing.

Types

WritingReviewEvalInput

What an eval hands the reviewer: the task the text was written for, and the file holding that text, seeded into the working directory by the test's files/. The shape the evals/writing-review suite uses as its input.

/** What an eval hands the reviewer: the task the text was written for, and
  the file holding that text, seeded into the working directory by the
  test's `files/`. The shape the `evals/writing-review` suite uses as its
  input. */
export type WritingReviewEvalInput = {
  assignment: string;
  sourceFile: string
}

(source)

Functions

buildTools

buildTools(): any[]

Return the writing reviewer's tools: the file tools (to read the text under review, and to apply fixes when asked), the shipped skills, and progress reporting.

Returns: any[]

(source)

writingReviewAgent

writingReviewAgent(
  text: string,
  task: string = "",
  guidelines: string = "",
  context: string = "",
  maxCost: number = $5.00,
  maxTime: number = 10m,
  model: string = "",
  provider: string = "",
  session: string = "",
  extraTools: any[] = [],
  failOpen: boolean = true,
): Result<Feedback[]>

Review prose for readability and return findings. error=true marks a passage a reader will misread or lose; error=false is advisory polish. Spelling and grammar are assumed checked separately: this agent reports only what needs judgment.

@param text - The text under review, or a request naming the files that hold it (and, if the caller wants the fixes applied, saying so) @param task - Who the text is for and what it must get across, or "" @param guidelines - The caller's own writing guidelines, as text (for example the contents of a project's writing-tips document). Outranks the built-in catalog where they conflict; "" for the built-in catalog alone @param context - Extra material for the judgment, or "" @param maxCost - Hard spend cap @param maxTime - Hard wall-clock cap @param model - Model override, or "" for the ambient model @param provider - Provider for the model override @param session - Session name to share a thread across calls, or "" for isolated @param extraTools - Extra tools to offer the LLM, appended to the built-in set @param failOpen - When true, a review that could not run (budget, policy, model error) returns no findings instead of a failure. Pass false to get the failure, for callers that must not mistake a failed review for a clean one

Parameters:

Name	Type	Default
text	string
task	string	""
guidelines	string	""
context	string	""
maxCost	number	$5.00
maxTime	number	10m
model	string	""
provider	string	""
session	string	""
extraTools	any[]	[]
failOpen	boolean	true

Returns: Result<Feedback[]>

Throws: std::guard

(source)