agentsclimarketplace

Persona review

Skill TaiSakuma/reviewed-writer/skills/persona-review

Claude Code plugin: persona-panel authoring and review workflow with Diátaxis declarationsFrom the repository description

Install
npx -y skills add TaiSakuma/reviewed-writer --skill persona-review

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

  • 1 stars1 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

SKILL.md

3.3 KB, as published. Nobody here has run it

Run one round of persona review: fixed-persona reviewers, each in its own subagent, review a document (or several drafts of it), and their reviews are collected and consolidated for the caller — an authoring workflow or a direct invocation.

Every repository-specific value lives in the repository's .claude/rules/persona-review-profile.md (the profile); read it first. The Diátaxis rules are the shared review core at ${CLAUDE_PLUGIN_ROOT}/skills/persona-review/references/diataxis-review.md; read it too — composing the brief needs its reader questions.

Steps

  1. Determine the review request. When invoked from an authoring run, the run state in the conversation supplies the draft path(s), the document's purpose, what is in and out of scope, the declared quadrant(s), each unit's status (when the profile's Status dimension section is enabled), the rubric, the verified facts and link targets, and the personas chosen for the run. When invoked standalone, default to: the shipped document named in the profile's Document section, as it stands; all personas listed in the profile's Personas section; implemented status, when the status dimension is enabled; no rubric.

  2. Compose the review brief to a temp file: the project and document identity, from the profile's Document section; the document's purpose; the declared quadrant(s) and matching reader question(s), carried as the profile's Declaration mechanism section directs; when the status dimension is enabled, each unit's status, and the design decisions for spec content — the brief is self-contained; what is in and out of scope; the rubric, when the request has one; verified facts; link targets; and the path of the Diátaxis core, so reviewers read the core from the brief. Reviewers never depend on files outside the repository, this skill, and the brief.

  3. Launch the panel. For each persona in the request, launch one reviewed-writer:persona-reviewer subagent via the Agent tool's subagent_type, all in parallel. Open each task prompt with the full content of that persona's head file, from the profile's Personas section, then give the brief path and the draft path(s). Ask each for: answers to the reader questions for the content its lens serves; a score per draft on the rubric axes, when the brief carries a rubric; lens-specific flags with quoted text and file:line citations; how relevant the document is to it (per section, when the section set is an output of the run); the best draft overall and per axis, when several drafts are under review; specific fixes; structural recommendations (sections to add, split, merge, or remove) when the section set is an output of the run; the single most important improvement; an alignment self-check per the Diátaxis rules (out-of-scope asks and their routing; out-of-quadrant content flagged); and a one-line ship/revise verdict (with the single most important change if revising).

  4. Collect and consolidate. If a reviewer errors out mid-run, re-launch it — do not treat a missing verdict as a pass. Consolidate the reviews into a matrix and report it to the caller with each reviewer's verdict and flags.

Keep looking

Skills are one crate of 608,666. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.