Application tracker

A PhD applicant is running many parallel processes — different deadlines, document checklists, and pending replies across a dozen labs — and the failure mode is silent: a missed deadline or a follow-up that never happened costs an entire opportunity. This skill is the applicant's air-traffic control. Its value is surfacing the few things that need action now out of a mass of state, accurately, so nothing winnable slips.

What you produce

An updated set of status.md files plus a concise action digest for the user: what's overdue, what's due soon, what's missing, and what's waiting on someone else. Accuracy and prioritization matter more than completeness — a digest that buries the one urgent deadline under thirty lines of "no action needed" has failed.

Step 1 — Gather state

Read every knowledge-base/applications/*/status.md (the stage, deadline, priority, checklist, and timeline) and the knowledge-base/interactions/*.md logs (what was sent, when, and any Follow-up due dates). Determine today's date with the shell rather than assuming it — overdue/upcoming calculations depend on it.

Step 2 — Compute what needs attention

Sort everything into a small number of action buckets, ordered by urgency:

• Overdue — a passed application deadline still marked unsubmitted, or a Follow-up due date in the past with status still "awaiting reply". These are the headline items.
• Due soon — deadlines or follow-ups within the next ~14 days. Show the date and the days remaining so urgency is legible.
• Missing documents — checklist items still pending for applications whose deadline is approaching (transcripts, letters, writing samples, test scores). Surface these early because documents like recommendation letters have their own lead time.
• Waiting on others — sent emails or submitted applications where the ball is in the professor's or committee's court; note when a polite follow-up becomes appropriate.
• No action needed — everything else, summarized in a single line, not enumerated.

For follow-up timing, respect the dates logged by outreach-email; a cold email typically warrants a follow-up ~2 weeks out, and only one or two follow-ups before moving on.

Step 3 — Update the records

Where the state has clearly changed, update the status.md files: advance stage, tick completed checklist items, correct deadlines you've confirmed. Keep edits faithful to what you actually know — don't mark something submitted on assumption. If you change a record, note it so the user can see what moved.

Step 4 — Deliver the digest

Lead with the action buckets, most urgent first. Be specific and scannable:

## PhD application status — <date>

### ⚠ Overdue / act now
- <App>: follow-up to Prof X is 5 days overdue (emailed <date>, no reply) → send follow-up
- ...

### Due within 14 days
- <App>: application deadline <date> (in 9 days); still missing: writing sample

### Waiting on others
- <App>: submitted <date>; committee decisions typically <timeframe>

### On track
- 6 other applications, nothing due in the next two weeks.

End with a short recommended next-action list tied to the urgent items (e.g. "draft the overdue follow-up to Prof X" → hand to outreach-email; "request the missing transcript").

Step 5 — Offer recurring tracking

Staying on top of deadlines is inherently recurring, and the whole point is to catch things before they're missed. Offer to set up a scheduled task (e.g. a Monday-morning digest) that re-runs this check and surfaces only what needs action that week. Many applicants don't realize this can run automatically.

Guardrails

Never invent or guess a deadline or a status — an applicant who relies on a fabricated date is worse off than one who knows the date is unconfirmed. When a date or requirement is unverified, say so and flag it for the user to confirm at the source. Don't auto-send follow-ups; surface that one is due and hand the drafting to outreach-email for the user to review and send.