From literate-agent
Audit one .org file's prose against the lp-prose-no-self-narration rule. Reports paragraphs that fail the deletion test (self-narration, filler parentheticals, mechanism-without-action). User-invocable only — do NOT auto-trigger on every edit.
How this skill is triggered — by the user, by Claude, or both
Slash command
/literate-agent:audit-prose <file-or-module>When to use
invoke as `/audit-prose <file.org>` or `/audit-prose <module-name>` when the user wants a focused prose-quality pass on one literate org module. Typical triggers: "audit the prose in X", "check this org file for filler", "lint the prose".
<file-or-module>**/*.orgThis skill is limited to the following tools:
The summary Claude sees in its skill listing — used to decide when to auto-load this skill
This is a **manual** companion to the `/docs-first` skill.
This is a manual companion to the /docs-first skill.
docs-first fires automatically whenever a literate .org file is
edited; /audit-prose is a deliberate one-shot pass over an
existing module to find dead prose. Two-skill split rationale:
auto-firing a prose-audit on every edit would flood context with
nits unrelated to the change in flight.
$ARGUMENTS — a file path or a module short name. If it's a path
ending in .org, use it directly. Otherwise glob from the current
project root for an .org file matching the short name (e.g.
process → look for *process*.org).
echo "Target file: $ARGUMENTS"
ls -la *"$ARGUMENTS"*.org 2>/dev/null | head -3
Apply the deletion test paragraph-by-paragraph. The canonical rule
lives at ~/projects/literate-agent/rules/lp-prose-no-self-narration.md
— consult it for the full taxonomy. The high-value failure modes:
Self-narration immediately after the artefact.
(Source: foo.svg — GitHub renders the SVG inline.) after an
org link. The link IS the source.The table above shows ... repeating the table content.This section provides ... restating the heading.Mechanism-without-action. Explaining how something works when the explanation neither warns the reader of a trap nor changes what they do next.
The :tangle path resolves relative to the .org file. →
delete. Only keep this if it carries a warning ("a git mv
of the .org silently breaks tangling").Filler parentheticals. (see below), (for details),
(note that ...), (as mentioned earlier) when the referent
is already in the reader's working memory.
Pleasantries & meta-narration. First, let's understand…,
In this section we will…, Hopefully this clarifies….
Restating the heading. A first sentence that just says what the heading already said. Either delete or merge into the heading.
#+END_SRC and the next
heading, or between a heading and the next #+BEGIN_SRC):
A table grouped by category, plus a one-line summary at the top:
audit-prose: <file.org> — N violations across M paragraphs
| Lines | Category | First 12 words | Suggested action |
|--------|----------|-----------------------------------------------|------------------|
| 42-45 | (1) self-narration | "(Source: architecture.svg — GitHub …)" | delete |
| 78-80 | (3) filler | "(See the next section for the routing logic.)" | delete |
| 120-128| (2) mechanism-no-action | "The :tangle path resolves relative to the .org" | delete OR add warning |
Run with --apply to land all "delete" suggestions in one commit (NOT
implemented in this version; manual apply for now).
The description + when_to_use frontmatter cover:
npx claudepluginhub jingtaozf/literate-agent --plugin literate-agentProvides behavioral guidelines to reduce common LLM coding mistakes, focusing on simplicity, surgical changes, assumption surfacing, and verifiable success criteria.
Searches, retrieves, and installs Agent Skills from prompts.chat registry using MCP tools like search_skills and get_skill. Activates for finding skills, browsing catalogs, or extending Claude.
Creates, edits, and optimizes skills for Claude Code, including drafting, evaluating with test prompts, iterating on performance, and improving skill descriptions for better triggering accuracy.