Critique Loop Writer

An iterative writing agent that researches, drafts, and self-critiques through a skeptical reader persona.

Install
cmdop skills install agensi-critique-loop-writer

Critique Loop Writer is a skill that runs a Draft → Critique → Revise cycle on every piece of writing it produces. Before returning output, it researches genre best practices and audience expectations in the background, then constructs a custom skeptical reader persona — for example, a cynical senior engineer, a weary literary agent, or a hostile-but-fair op-ed reader — to evaluate the draft. The agent revises iteratively until that internal reviewer has no substantive objections remaining.

The process specifically targets recognizable AI writing patterns: overused words and phrases such as “delve,” “tapestry,” and “in today’s fast-paced world,” as well as uniform sentence rhythms that make prose feel machine-generated. Because the skeptical persona reflects the biases of the intended audience, the critique is not purely stylistic — it also checks whether arguments hold up to real-world scrutiny from that audience’s perspective.

Output is delivered in Markdown or plain text, described as ready for publication. Supported genres include essays, technical blog posts, speeches, op-eds, creative writing, and personal statements. The skill is explicitly not intended for brief emails or cover letters.

There are no exposed tools and no listed environment variables, so this skill operates as a self-contained agentic loop rather than a configurable pipeline with external integrations.

Use cases

  • Draft and self-critique a technical blog post until it passes review by a simulated skeptical senior engineer persona
  • Generate an op-ed that survives adversarial review targeting weak arguments and generic phrasing
  • Produce a personal statement that has been iterated against audience-specific biases before delivery
  • Write a speech or essay with metronomic sentence rhythms identified and corrected internally
  • Remove common AI writing tells from long-form prose before publication

When to use it

  • When producing long-form content such as essays, technical posts, op-eds, speeches, or personal statements
  • When the output must not read as AI-generated and common clichés need to be actively hunted down
  • When argument quality and audience reception matter as much as surface-level grammar
  • When a multi-round internal revision loop is preferable to a single-pass prompt

When not to use it

  • For brief emails or cover letters, which the skill explicitly does not target
  • When a fast, single-pass draft is sufficient and iterative revision latency is unacceptable
  • When the required genre falls outside essays, blog posts, speeches, op-eds, creative writing, or personal statements
  • When external tool integrations or configurable environment variables are needed, as none are provided