aeroplan-activity

Unified output contract. This sub-skill is part of points-activity. Its scripts/transform.py writes Date, Description, Amount and the filename aeroplan_activity_<from>_<to>.csv (covered range) via the shared scripts/activity_output.py, and prints BALANCE:/COVERED:/REQUESTED:/FILE:/ROWS: to stdout. The orchestrator passes the requested period and reads those lines. References to collapsed.csv / Points / Miles below are the older single-program wording; the actual columns are Date, Description, Amount.

Air Canada Aeroplan Activity Extraction

Scrapes the rendered Aeroplan activity dashboard, sets the period filter to its max (2 years), classifies and collapses, and outputs a points table.

Prerequisites

• Claude in Chrome extension connected (list_connected_browsers → select_browser).
• User logged in to aircanada.com. Don't log in on their behalf.
• Verify login programmatically: a points balance shows top-right (e.g. "124,520 pts") and there are no sign-in buttons.

Why DOM scraping (not the API)

Activity is served by an AWS AppSync GraphQL endpoint (akamai-gw.dbaas.aircanada.com/appsync/transaction-history-v2). A plain fetch fails (CORS / needs the Amplify client's auth), and the real request doesn't pass through window.fetch/XHR so it can't be intercepted. The data renders into the normal light DOM (Angular Material mat-expansion-panel), so scraping is the reliable path.

⚠️ Family Sharing caveat (read this)

If the account is a Family Sharing pool lead, the visible balance INCLUDES other pool members, but the activity list shows only THIS account's transactions. Other members' transfers/earns/returns are NOT here (e.g. a spouse's Chase transfer and some return credits won't appear), even though they affect the shared balance. So: do not reconcile the activity total to the balance — the mismatch is structural, not an extraction error. Treat the output as "activity visible on this account".

Procedure

1. Open the page

tabs_context_mcp(createIfEmpty:true) → navigate to https://www.aircanada.com/aeroplan/member/dashboard/activity. Confirm login. Decline the cookie banner ("Reject All") if present. Keep tabId.

2. Scrape (sets 2-year filter automatically)

Run scripts/scrape_activity.js. It opens Filters, selects "2 years" (the max offered — 30/60/90 days, 6 months, 1 year, 2 years), applies, clicks any "More"/"Load more" until the count stabilizes, then parses each <mat-expansion-panel> header. Ad panels (e.g. "Earn points with Starbucks") have no "Pts" and are skipped. Returns {count}, stashes rows on window.__aero.

Note: expanding a panel reveals a your-balance / other-pool-member split — we IGNORE that and take only the header total (per requirements).

3. Dump and read back

Run scripts/dump_console.js (logs AE###~ ...), then read_console_messages(pattern:"^AE\\d", limit:60, tabId:...).

4. Transform

Strip the AE###: prefix, write the |-delimited lines to a file (e.g. /tmp/raw.txt), then:

python3 scripts/transform.py /tmp/raw.txt <output_dir> [from] [to] [balance]
# from/to are ISO yyyy-mm-dd or '-' for unbounded; balance is the current balance or omit.

Writes collapsed.csv, prints row count + points total.

5. Present

present_files on collapsed.csv. Don't reconcile to balance (see Family Sharing).

The collapsed table (the deliverable)

collapsed.csv: columns Date, Description, Points. Logic (in scripts/transform.py):

1. Classify: redemption = Flight Reward Booking, its Cancellation, or an Aeroplan Family Sharing redemption (tied to a booking, matched to itineraries later); everything else (card transfers like Amex / Chase Ultimate Rewards, bonuses, partner earns) is an earning/transfer.
2. Redemptions: keep the real transaction date.
3. Earnings/transfers: move the date to the last day of its month.
4. Collapse by (Date, "Partner: Description"), sum Points. Identical rows on the same date merge (e.g. two equal Flight Reward Bookings → one summed row); rows that differ (e.g. cancellations carrying distinct transaction numbers) stay separate.
5. Drop zero-point rows after collapsing.

Recurring / scheduled use (monthly)

Run as a scheduled task on the user's own machine — needs the user's real Chrome + live Aeroplan session. Use the host's scheduling primitive: Cowork — mcp__scheduled-tasks__create_scheduled_task; Claude Code — cron + headless invocation. Caveats: local tasks run only while the computer is on and the app is open; no completion/failure notifications yet; an expired session needs a fresh login. Each run is a clean session — to report only new activity, save each run's collapsed.csv (dated filename) and diff against the prior one. Note the 2-year window: anything older than 2 years can never be retrieved, so a monthly cadence is more than enough to never miss a transaction.