Refreshes README structure favoring concision (an index, not a manual). Use when the README needs a structural update after significant project changes.
Standardmäßig ist der Prompt ausgewählt, der zuerst die Quelle prüft. Sie können zu einem direkten Befehl wechseln oder eine lokale Kopie herunterladen.
Quelldateien prüfen
Lesen Sie SKILL.md und alle von SkillsMP angezeigten Begleitdateien, bevor Sie sich für eine Installation entscheiden.
Mit Codex oder Claude installieren Kopieren Sie diesen Prompt, fügen Sie ihn in Codex, Claude oder einen anderen Assistant ein und lassen Sie die Skill-Seite prüfen und installieren.
Ein direkter Befehl überspringt den Prüf-Prompt. Prüfen Sie die Quelle, bevor Sie ihn ausführen.
Refreshes README structure favoring concision (an index, not a manual). Use when the README needs a structural update after significant project changes.
Use this skill whenever the README requires a structural refresh.
Run Skill(sanctum:git-workspace-review) first to capture repo context and diffs.
When NOT To Use
Updating inline docs: use doc-updates
Consolidating ephemeral reports: use doc-consolidation
Required TodoWrite Items
update-readme:language-audit
update-readme:exemplar-research
update-readme:outline-aligned
update-readme:edits-applied
update-readme:slop-scanned - AI marker detection via scribe
update-readme:verification-reporting
Step 1 - Language Audit (update-readme:language-audit)
Confirm pwd, git status -sb, and the baseline branch for reference.
Detect dominant languages using repository heuristics (manifest files, file counts).
Note secondary languages that influence documentation (e.g., a TypeScript frontend and a Rust backend) so the README can surface both.
Record the method and findings.
See modules/language-audit.md for detailed detection patterns and commands.
Step 2 - Exemplar Research (update-readme:exemplar-research)
For each primary and secondary language, use web search to locate high-quality READMEs (star count, recency, maintainer activity).
Capture 2-3 exemplar repositories per language and summarize why each is relevant (section order, visuals, quickstart clarity, governance messaging, math exposition, etc.).
Store citations for every exemplar so the final summary references them explicitly.
See modules/exemplar-research.md for search query patterns and evaluation criteria.
The README is an index, not a manual. Lead with what the
project is, how to install it, and a scannable table of its
units (plugins, commands, packages); link out for everything
else. When this repo's root README was simplified in 2026, it
dropped from 435 to 200 lines (65% fewer words) by deleting
inlined depth and linking to book/ instead. Aim for that
shape, not a section-by-section coverage sweep.
Compare current README headings ()
against the exemplars from Step 2. Lean marketplace and
CLI-tool READMEs (asdf-vm/asdf-plugins, ohmyzsh/ohmyzsh,
ryoppippi/ccusage) are index-shaped: prose intro under ~500
words, a unit table doing the work, depth behind links.
rg -n '^#' README.md
Keep only the sections that earn a place in an index: a
one-line value proposition, one install path, a compact
command or unit table, a short "what's inside" summary, a
single Documentation link section, and metadata
(requirements, contributing, license) last. For every other
candidate section ask: does this belong in the README, or
behind a link?
Default to linking out. Send deep architecture, FAQ,
roadmap, changelog, alternate install methods, and per-unit
detail to their own pages. Per-unit docs live with the unit
(hub-and-spoke), never duplicated in the root.
Order signals priority: title and tagline first, then
install, then the index table, then a short feature summary,
then Documentation links, with housekeeping (requirements,
contributing, license) last.
Map internal documents (docs/, book/, specs/, commands/) to
the section that links them, so the README anchors depth
without restating it.
Soft budget: a marketplace or collection root README runs
roughly 150-400 lines of prose plus any index table. A section
that grows over time (roadmap, FAQ, changelog) belongs in a
linked page, not the README.
Implement the new structure directly in README.md
(or the specified file).
Follow Skill(leyline:markdown-formatting) conventions:
wrap prose at 80 chars (prefer sentence/clause boundaries),
blank lines around headings, ATX headings only, blank line
before lists, reference-style links for long URLs.
Prefer deletion and linking over inlining. If a section
duplicates a docs page, cut it and link the page; if it
grows over time, move it out of the README entirely.
Add comparison tables, feature lists, or diagrams only if
they originate from current repository assets (no speculative
content).
When referencing algorithms or performance claims, point to
benchmarks or tests within the repository or documented math
reviews.
Step 4.5 - AI Slop Detection (update-readme:slop-scanned)
Run Skill(scribe:slop-detector) on the updated README to detect AI-generated content markers.
Scribe Integration
The scribe plugin provides AI slop detection:
Skill(scribe:slop-detector) --target README.md
This detects:
Tier 1 words: delve, tapestry, comprehensive, leveraging, etc.
Phrase patterns: "In today's fast-paced world", "cannot be overstated"
Structural markers: Excessive em dashes, bullet overuse, sentence uniformity
Re-read the updated README for clarity, accessibility (section lengths, bullet balance), and accurate links.
Run git diff README.md (or the edited file) and capture snippets for the final report.
Summarize detected languages, exemplar sources (with citations), key structural decisions, and follow-up TODOs (e.g., add badges, upload diagrams).
Exit Criteria
All TodoWrite items are complete.
The README is index-shaped: prose intro within the soft
budget (roughly 150-400 lines plus any unit table), one
install path, and a scannable unit/command table.
No README section duplicates a book/ or docs/ page;
depth is linked, not inlined (rg -n '^#' README.md headings
each map to a section that earns a place in an index).
The README references internal docs and external
exemplars with citations.
Research notes and command references are captured so
future reviewers can reproduce the process.
Troubleshooting
Common Issues
Documentation out of sync
Run make docs-update to regenerate from code
Build failures
Check that all required dependencies are installed
Links broken
Verify relative paths in documentation files
