← ClaudeAtlas

project-review-docslisted

Read-only audit of a project's docs for accuracy, staleness, gaps, misplaced content, and whether agents can and do actually use them — runs a multi-agent workflow, reports fixes, never edits.
hk9890/claude-dev-tools · ★ 0 · Code & Development · score 76
Install: claude install-skill hk9890/claude-dev-tools
Read-only documentation audit. Launch the review workflow — do **not** review the docs inline. The workflow returns a structured report; relay it. ## Run the workflow 1. Parse `$ARGUMENTS` as `[low|medium|high|ultra] [what-to-review]`. Both are optional. A leading `low` | `medium` | `high` | `ultra` token is the **level** (default `medium`); everything after it is **what to review**. Most invocations pass only a level. Unlike the other reviewers, what-to-review here must resolve to a **path** — `manifest.py` takes a directory, not a free-form description. Default: the repo root. If the argument is a description rather than a path, resolve it to a directory or fall back to the root. 2. Load the standard. `SKILL_DIR` is the **base directory for this skill**, given at the top of this file when the skill loads. It is absolute and install-correct — build every path below from it. The standard this review measures against is **not** in this plugin: the ownership contracts and the authoring rules belong to `instruction-writing:writing-project-docs`. Load that skill now — the same "Base directory for this skill" line comes with it, and that path is `STANDARD_DIR`. Take it from the load; never construct it and never search the filesystem for the plugin: a sibling plugin's install directory is not derivable from this one's, and a guess selects a stale cached version. If the skill is not installed, stop and say so — `project-review` d