R Weekly Highlights Poll Generator

Replicates the rweekly.highlights Shiny app (https://github.com/rweekly/rweekly.highlights) as a Claude Code skill.

Goal

Produce two Slack /poll commands for the #highlights channel — one for items 1–5, one for items 6–10 — so editors can vote on the weekly highlights.

Steps

Step 1: Parse the draft

Read draft.md and extract the issue number from the title: front matter field (e.g. R Weekly 2026-W24 → 2026-W24).

If the title contains the placeholder W00, determine the correct week number by:

1. Running ls _posts/ | sort | tail -1 to find the last published issue filename (e.g. 2026-06-08-2026-W24.md)
2. Extracting its week number and adding 1 (e.g. W24 → W25)
3. Cross-checking with date +%G-W%V — use whichever is later

Use the resolved week number (e.g. 2026-W25) everywhere ISSUE appears in the poll commands.

Extract every link in the draft body using the pattern + [Title](URL). Exclude:

• Lines inside the ### Highlight section (already selected as highlights; note that highlight links are copied — they also remain in their original sections, so the same URL may appear twice in the draft)
• Boilerplate/recurring links:
  • https://serve.podhome.fm/r-weekly-highlights
  • https://rweekly.org/live
  • https://jumpingrivers.github.io/meetingsR/events.html
  • https://community.rstudio.com/c/irl
  • https://sites.google.com/view/dariia-mykhailyshyna/main/r-workshops-for-ukraine
  • https://DSLC.io/
  • https://github.com/rweekly/rweekly.org#how-to-have-my-content-shared-by-r-weekly
  • https://r-universe.dev/search/
  • https://dirk.eddelbuettel.com/cranberries/cran/new/

Display the full numbered list to the editor, grouped by section, so they can see where each item comes from.

Before displaying the list, run the duplicate checker against the last 20 published posts:

Rscript -e 'source("scripts/find_duplicates.R"); dups <- get_dups(); if (!is.null(dups)) cat(dups, sep="\n")'

Mark any link that appears in the output with a [DUP] label in the numbered list so the editor can see it at a glance. Duplicates are not automatically excluded — some repeats are intentional (events, conferences, recurring resources) — but they should generally be deprioritised when suggesting highlights.

Note: get_dups() also warns about within-draft duplicates. Ignore those warnings for links in the ### Highlight section — those are intentional copies of links from their original sections.

Step 2: Pre-select candidates

Review the list and suggest 10 items that would make strong highlights — prioritise:

• Broad appeal to R users of all levels
• Tutorials and practical tips
• Notable package releases or updates
• Community news from rOpenSci, Posit, etc.
• Anything particularly novel or creative

Before presenting your picks, re-verify each of the 10 by fetching its actual page content — draft.md is expected to already be R-related content, but curation mistakes happen, and a highlight gets far more visibility than a regular draft entry (editor poll → newsletter highlight). If a pick turns out not to be genuinely R-related on inspection, do not silently drop and replace it — tell the editor which item failed verification and why, propose a specific replacement, and let them confirm before it goes into the list you present.

Present your 10 suggested picks clearly (e.g. "My suggested 10: items 3, 7, 12, ...") and ask the editor to confirm or swap any out.

Wait for the editor's response before proceeding.

Step 3: Generate Slack poll commands

Once exactly 10 items are confirmed, format two Slack /poll commands.

Each item is formatted as:

"TITLE\nURL"

Poll 1 (items 1–5):

/poll "Which of the following items in ISSUE should be highlighted? (Part 1)" "Title1\nURL1" "Title2\nURL2" "Title3\nURL3" "Title4\nURL4" "Title5\nURL5"

Poll 2 (items 6–10):

/poll "Which of the following items in ISSUE should be highlighted? (Part 2)" "Title6\nURL6" "Title7\nURL7" "Title8\nURL8" "Title9\nURL9" "Title10\nURL10"

Output both commands as plain text in a code block, ready to copy and paste into the #highlights Slack channel.