likutei-peshatim

Parse a Likutei Peshatim (HTC weekly Torah newsletter) email into calendar events and publish them to /calendars. Use when the user says /likutei-peshatim, "parse this week's Likutei Peshatim", or nam

SKILL.md

GitHub

Likutei Peshatim — newsletter to calendar

Every week HTC emails "Likutei Peshatim: <Parsha> - <Year>" to matthewjmiller07@gmail.com. It's a dense community bulletin (Torah insights, shul announcements, job postings, yahrzeits, event listings) mixed together in one wall of text. This skill extracts the calendar-worthy events and appends them to src/data/likutei-peshatim/events.json, which scripts/likutei-peshatim/generate-ics.mjs turns into public/calendar/likutei_peshatim.ics, served from the /calendars page.

1. Get the raw email

Automated runs pass a scratch text file path. Two runners write it, with the same output and the same processed-emails.json bookkeeping, so whichever runs first wins: .github/workflows/likutei-peshatim.yml (the main one: fetch-new-emails-imap.py reads the inbox over IMAP with a Gmail app password; Thursday night, plus a staleness alarm) and the Mac's launchd weekly-run.sh (fetch-new-emails.mjs, via the local gws CLI). For an ad hoc run on a specific message, fetch it directly:

gws gmail users messages get --params '{"userId":"me","id":"<messageId>","format":"full"}'

The plaintext body is at payload.parts[].body.data (or nested deeper for multipart/mixed) — base64url-decode it.

First check src/data/likutei-peshatim/processed-emails.json — if this message ID is already there, stop; it's already been parsed. Never reprocess.

2. What counts as an event

Include anything with a resolvable date: shiurim, drashos, screenings, races, fast-day schedules, fundraising campaigns with a date window, Scholar-in-Residence talks, and yahrzeits (as all-day events — Matthew has explicitly asked for these to be included).

Skip: job postings, open-ended asks ("join our waitlist", "seeking a...for hire"), and anything with no date at all.

Granularity: prefer one event per distinct time/session rather than collapsing a whole schedule into one block (e.g. a Tisha B'Av schedule with 4 service times → 4 events, not 1). When several locations share one time (e.g. "NILI Shabbat shiurim, all at 5:30 PM" across 3 homes), one event with all locations/speakers in the description is fine.

3. Resolving dates

The email header always anchors a Hebrew date to a Gregorian one, e.g. 2 Av 5786 | July 16, 2026. Most events already give an explicit Gregorian date in the body ("Monday evening, July 20th") — use that directly.

For Hebrew-only dates (yahrzeits are almost always given only as "7 Av", "כ״ח תמוז", etc.): never hand-compute the Gregorian date from month-length knowledge or day-offset arithmetic — always resolve it with the Hebcal converter API. The Hebrew year comes from the email header (e.g. 5786). Call:

https://www.hebcal.com/converter?cfg=json&hy=<hebrew_year>&hm=<hebrew_month>&hd=<hebrew_day>&h2g=1

hm takes standard English transliterations (Tamuz, Av, Elul, Tishrei, Cheshvan, Kislev, Teves, Shvat, Adar, Adar1/Adar2 in a leap year, etc. — not the Hebrew glyphs from the email). The response's gy/gm/gd is the Gregorian date. Example:

curl -s "https://www.hebcal.com/converter?cfg=json&hy=5786&hm=Av&hd=2&h2g=1"
# {"gy":2026,"gm":7,"gd":16,...}  →  2026-07-16

For Hebrew-only dates written in Hebrew glyphs (e.g. "כ״ח תמוז"), transliterate the month and convert the gematria day number before calling the API — don't guess the Gregorian date from the glyphs directly.

Still sanity-check the result against known fixed points stated elsewhere in the same email (e.g. Tisha B'Av) and against day-of-week where given ("Wednesday 8/22/26" that doesn't match the API's date is very likely a copy-paste typo in the newsletter itself — trust the API, note the correction in the event's description, don't blindly transcribe an inconsistent date).

Times are America/Chicago local time — store as plain "HH:MM" (24h), not with an offset; generate-ics.mjs handles the UTC/DST conversion.

4. Event schema

Append to src/data/likutei-peshatim/events.json (array), each entry:

{
  "id": "lp-YYYYMMDD-short-slug",
  "title": "string",
  "date": "YYYY-MM-DD",
  "time": "HH:MM",            // omit or set allDay:true for all-day (yahrzeits, etc.)
  "endTime": "HH:MM",         // optional; else durationMinutes (default 60) is used
  "durationMinutes": 60,      // optional
  "endDate": "YYYY-MM-DD",    // optional, for multi-day all-day spans
  "allDay": false,
  "location": "string or null",
  "description": "string",
  "sourceExcerpt": "verbatim quote from the email this event was parsed from",
  "category": "shiur|drasha|screening|race|fast|campaign|event|yahrzeit",
  "sourceMessageId": "<gmail message id>",
  "sourceSubject": "<email subject>",
  "sourceDate": "YYYY-MM-DD"  // date the email was sent
}

id must be unique and stable — dedupe against existing entries in events.json (same title+date already present from a prior week means don't re-add).

5. Finish: mark processed, rebuild, ship

  1. Append {id, subject, date, eventCount} to src/data/likutei-peshatim/processed-emails.json for this message — even if eventCount is 0 (e.g. an addendum with no dated content), so it's never reprocessed.
  2. Run node scripts/likutei-peshatim/generate-ics.mjs to rebuild the ICS file.
  3. git add only the 3 files touched (events.json, processed-emails.json, public/calendar/likutei_peshatim.ics) plus calendars.astro if the event count line needs no edit (it's computed from events.json automatically — don't touch calendars.astro for routine runs).
  4. Commit (Add Likutei Peshatim events: <Parsha> - <date>) and push to origin main.