Save valuable PR review comments as markdown. AI agents load them before writing code, so past feedback isn't repeated. PR Review Comment → Save as Markdown → AI Agents Learn → Better Code
npx skills add damengrandom/learn-from-pr -g # global
npx skills add damengrandom/learn-from-pr # this project only/skill:learn-from-pr save "Avoid watch for derived state. Use computed — it's cached." \
--title "prefer-computed-over-watch" \
--tags "vue,reactivity" \
--image "./before.png|Broken layout"OR
# Tell Claude code, please run command:
/learn-from-pr save --title "kebab-case-slug" --tags "performance" --image [#image1], [#image2]
# Example
# /learn-from-pr save Use $slots checks when wrapper elements add spacing, borders, or layout constraints. --title "Conditionally Render Optional Slot Wrappers" --image [Image #1] [Image #2] --tag "slots, vue, component"| Option | Description |
|---|---|
<comment> |
The review text, quoted exactly as written (required) |
--title "slug" |
Kebab-case filename; the heading derives from it (prefer-computed-over-watch → # Prefer Computed Over Watch) |
--tags "a,b" |
Comma-separated |
--image "path|alt" |
Repeatable. Paths must be absolute or relative to where you run it |
/skill:learn-from-pr list
# Claude code
/learn-from-pr listTitles, dates, and tags for every saved lesson.
/skill:learn-from-pr remove "prefer-computed-over-watch"
# Claude code
/learn-from-pr remove "prefer-computed-over-watch"Delete a lesson by slug.
/skill:learn-from-pr uninstall
# Claude code
/learn-from-pr uninstall
⚠️ Deletes the whole skill folder immediately, lessons and images included. No confirmation, no recovery.
references/ for the markdown, assets/ for the images, both inside the skill
folder.
⚠️ Commit them to git.npx skills addoverwrites the skill folder on every update, and git is what brings your lessons back. It's also how you share them with your team.
MIT