ihg-activity - SKILL.md Agent Skill

name: ihg-activity description: Extract IHG One Rewards account points activity into the unified points-activity CSV (Date, Description, Amount) plus the current balance. Use whenever the user wants to pull, export, check, sync, or monitor their IHG points history, account activity, stays, reward-night redemptions, or credit-card points — including recurring/scheduled monthly checks. Triggers on "get my IHG points activity", "update my IHG sheet", "IHG account activity". Part of the points-activity skill suite; scrapes the rendered activity table (~365 days). Requires the user to be logged in to ihg.com in the connected Chrome browser.

Unified output contract. This sub-skill is part of points-activity. Its scripts/transform.py writes Date, Description, Amount and the filename ihg_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.

IHG One Rewards Account Activity Extraction

Scrapes the rendered IHG account-activity table, classifies and collapses it, and outputs a points table. IHG only exposes ~~365 days (~~30 rows) of activity, and that is the full available history — there is no deeper archive to page through.

Prerequisites

Claude in Chrome extension connected (list_connected_browsers → select_browser).
User logged in to ihg.com. IHG sessions drop quickly and redirect to a sign-in page; the skill must not log in on the user's behalf.
Verify login programmatically: the activity page shows the point balance (e.g. "161,045 pts") and no sign-in buttons. If the URL contains /sign-in, stop and ask the user to log in.

Why DOM scraping (not the API)

The data API apis.ihg.com/members/v1/profiles/me/activities needs an apikey (in window.APIGEE_APIKEY_VALUE) plus the X-IHG-SSO-TOKEN header set; a plain fetch reproducing them still returns 504, and the Angular app doesn't use window.fetch/patchable XHR so the live request can't be intercepted reliably. Since the page only ever shows ~365 days anyway, scraping the rendered table is simpler and complete. (Endpoint shape, if ever revisited: ?activityType=all&duration=365&limit=10&offset=N, 0-based offset.)

Procedure

1. Open the page

tabs_context_mcp(createIfEmpty:true) → navigate to https://www.ihg.com/rewardsclub/us/en/account-mgmt/activity. Confirm login (see above). Keep the tabId.

2. Scrape

Run scripts/scrape_activity.js via javascript_tool (action javascript_exec). It scrolls to load all rows, parses each table <tr>, returns {count, newest, oldest}, and stashes rows on window.__ihg.

If a previous run left a foreign browser-extension node in the page, a global querySelectorAll can throw "Cannot access a chrome-extension:// URL". If that happens, re-navigate to the activity URL (fresh load clears it) and re-run.

3. Dump and read back

Run scripts/dump_console.js (logs IX###~ ... lines), then read_console_messages(pattern:"^IX\\d", limit:60, tabId:...). Direct JSON returns are often blocked by the chat ("Cookie/query string data"); the console path avoids it.

4. Transform

Strip the IX###: prefix from each line, write the remaining ~~-delimited lines to a file (e.g. /tmp/raw.txt), then run:

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.

It writes collapsed.csv and prints row count + total.

5. Present

present_files on collapsed.csv. Do NOT reconcile against the balance — the activity sum will not equal it (only ~365 days are shown). That mismatch is expected.

The collapsed table (the deliverable)

collapsed.csv: columns Date, Description, Points. Logic (all in scripts/transform.py — read its docstring before editing):

Classify each row as a redemption ("Redeemed points for Reward Night stay on MM/DD/YYYY ...") or an earning/transfer (everything else).
Redemptions: keep the real transaction date; strip the per-night date from the description so all nights of one booking collapse into ONE Stay row (a 4-night reward booking 0 / -45,000 / -42,000 / -45,000 → one -132,000 row). Real date kept because redemptions get matched to invoices later.
Earnings/transfers: keep the real transaction date, kind='earn'.
Emit 4-tuples; the shared activity_output.py collapses earnings by (Date, normalized Description), summing Points. Brand + hotel after ' | ' are KEPT (used to match reservations later).
Drop zero-point rows — but only AFTER collapsing, so a 0-point free night still merges into its booking first.

Recurring / scheduled use (monthly)

Run as a scheduled task on the user's own machine — needs the user's real Chrome + live IHG session, which only exist locally. Use the host's scheduling primitive: Cowork — mcp__scheduled-tasks__create_scheduled_task; Claude Code — cron + headless invocation. Caveats for the user: local tasks run only while the computer is on and the app is open; no completion/failure notifications yet; a month-old IHG session will almost always need a fresh login first. Each run is a clean session — to report only new activity, save each run's collapsed.csv (dated filename) and diff against the previous one.