agentleFS
quarto-devquarto-dev/q2/.claude/skills/weekly-summary/SKILL.md

weekly-summary

Skill302 starsChanged 2 months agoMITView on GitHub

Write a Slack-ready summary of Quarto 2 (q2) work landed on main since a point in the past — a short themes paragraph plus a changelog-style bullet list of merged PRs. Use when the user says "weekly summary", "what landed this week", "summarize recent work for Slack", or invokes /weekly-summary [since]. The optional argument is a date (2026-10-27), a relative span ("7 days", "2 weeks"), or a git ref/tag; the default is 7 days.

6 other projects use a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

4 rules. 0 are common, 3 are distinctive, and 1 is very specific to this project.

  • 0

    Common

    Most of the 3 edited versions we read keep these.

  • 3

    Distinctive

    This author's own choices. Worth a look.

  • 1

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Read-only: this skill fetches and reads.

    line 12
  2. Distinctive:

    If the result count hits the limit, say so and narrow the window rather than silently truncating.

    line 48
  3. Very specific:

    When a body is empty and the title is cryptic, read the diffstat (gh pr view N --json files --jq '.files[].path') rather than guessing.

    line 92
  4. Distinctive:

    Keep each bullet to one line where possible.

    line 127

“Common” is measured: kept by at least 6 in 10 of the 3 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Weekly summary

Produces a message the user will paste into Slack and lightly edit. The audience is the Quarto team and adjacent colleagues: people who know what Quarto is but did not follow every PR. Write for them, not for the PR authors.

Read-only: this skill fetches and reads; it never commits, pushes, comments, or posts anywhere.

1. Resolve the window

Parse the argument into a SINCE date (UTC, YYYY-MM-DD):

  • empty → today minus 7 days
  • YYYY-MM-DD → as given
  • N days / N weeks → today minus that span
  • anything else → treat as a git ref; SINCE is that commit's committer date (git log -1 --format=%cs <ref>)

State the window in one line before gathering ("Summarizing main from 2026-10-27 to 2026-11-03").

2. Gather

Fetch first — the local main is often behind, and the summary must reflect origin/main:

git fetch origin main --quiet

Merged PRs into main (the backbone of the summary):

gh pr list --repo quarto-dev/q2 --state merged --base main \
  --search "merged:>=$SINCE" --limit 300 \
  --json number,title,author,mergedAt,headRefName,labels,additions,deletions,body \
  --jq '.[] | {number, title, author: .author.login, mergedAt, head: .headRefName,
               labels: [.labels[].name], size: (.additions + .deletions),
               body: (.body // "" | .[0:800])}'

If the result count hits the limit, say so and narrow the window rather than silently truncating.

Direct commits to main (pushes that bypassed a PR — usually docs, skills, small fixes): first-parent, non-merge commits in the window.

git log origin/main --first-parent --no-merges --since="$SINCE" \
  --format='%h%x09%an%x09%s'

Drop any whose subject ends in (#NNN) — those are squash-merged PRs already counted above.

Integration branches. PRs merged into a feature/* or refactor/* branch (not main) are not shipped yet. Don't list them individually; when the integration branch itself lands on main, its PR covers them. If one is notably active, a single "in progress" line in the themes paragraph is enough:

gh pr list --repo quarto-dev/q2 --state merged --search "merged:>=$SINCE -base:main" \
  --limit 100 --json number,title,baseRefName

3. Understand before writing

Titles alone mislead. For each PR, use the title, branch name, and body excerpt to decide:

  • What changed for a user (a Quarto author, a hub-client user, an extension author) — or, for internal work, what it enables.
  • Area — pick from this list, adding one only if a cluster clearly doesn't fit:
    • Rendering & formats (HTML, PDF/Typst/LaTeX, docx, revealjs, websites, books, listings)
    • Authoring features (crossrefs, citations, shortcodes, callouts, brand, layout)
    • Execution & engines (knitr, jupyter, execute: options)
    • Extensions & Lua filters
    • Preview & hub-client (q2 preview, quarto-hub web app, collaboration, MCP)
    • Parser & language (tree-sitter-qmd, pampa, diagnostics)
    • Q1 compatibility — use when the PR's point is matching Quarto 1 behavior
    • Infrastructure (CI, build, deps, perf, test harness, agent tooling/skills)
  • Weight — new capability > user-visible fix > internal cleanup. Size (additions + deletions) is a weak hint, not a ranking.

When a body is empty and the title is cryptic, read the diffstat (gh pr view N --json files --jq '.files[].path') rather than guessing. Do this sparingly; most PRs are clear from title + branch + body.

Collapse noise: dependency bumps (Snyk, Dependabot, upgrade-cargo-deps), CI-only tweaks, and lint passes collapse into one bullet per kind ("Dependency upkeep: 4 PRs (Snyk npm bumps, cargo patch upgrades)").

4. Write

Two parts.

Themes paragraph (3–6 sentences). Lead with the biggest story of the period, then the next one or two. Name concrete outcomes ("q2 render now honors --no-execute and execute: under format:"), not activity ("lots of work on execution"). If there is no real theme, say the period was mostly fixes across areas and name the two or three most notable — don't invent a narrative. End with one sentence of totals: "N PRs merged, M direct commits, K contributors."

Changelog, grouped under the areas from step 3 (omit empty areas; order areas by how much landed). One bullet per PR:

- Short user-facing description of the change. (#1046)
  • Rewrite titles into plain past-tense or present-tense statements; drop conventional-commit prefixes, braid ids, and branch jargon.
  • Mention the author only when it's not the user (gh api user --jq .login) — e.g. "… (#1031, @someone)". Never @-mention bots.
  • Merge two or three PRs into one bullet when they are clearly one change delivered in pieces: "(#1040, #1041)".
  • Direct commits that matter get bullets too, with the short hash instead of a PR number; trivial ones (typos, notes) are omitted.
  • Keep each bullet to one line where possible.

Slack formatting

The user pastes this into Slack's composer, so:

  • Bold area headings with single asterisks: *Rendering & formats*.
  • Plain - bullets, no nesting.
  • PR references as bare #1046. If the user wants links, switch to full URLs (https://github.com/quarto-dev/q2/pull/1046); Slack's composer does not turn [text](url) markdown into links reliably.
  • No # headings, no tables.

Print the finished message inside a single fenced block (```text) so it copies cleanly, then, outside the block, a short list of judgment calls the user may want to revisit (PRs you weren't sure how to describe, things you grouped or dropped).

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.