skymap-sync-credits

name	skymap.sync-credits
description	Sync the Buy Me a Coffee sponsor list and GitHub contributor list into Sky Map v2's contributors.xml. Trigger on "sync sponsors", "update donors", "sync contributors", "refresh credits" for stardroid-v2.

Sky Map v2 Credits Sync

Updates sponsors_text and contributors_text in app/src/main/res/values/contributors.xml, mirroring stardroid-v1's skymap.sponsors / skymap.contributors skills but adapted for v2's merged file and the current (MCP-based) Buy Me a Coffee integration.

Both strings are translatable="false" and are spliced into credits_text (app/src/main/res/values/credits.xml) via stringResource format args in HelpScreen.kt — see AGENTS.md.

Contributors (GitHub) — fully scriptable

v1's no-checkin-bmac_pat-style GitHub token isn't needed; use gh auth token:

GITHUB_TOKEN="$(gh auth token)" python3 tools/sync-credits.py contributors

This fetches commit authors from sky-map-team/stardroid (shared repo, so it picks up v1 contributors too), resolves each login to a display name, dedupes, and rewrites contributors_text in place. Review the diff — GitHub logins without a public display name fall back to the raw login string.

Sponsors (Buy Me a Coffee) — requires the MCP server, not a script

Buy Me a Coffee retired the personal-access-token REST API that v1's tools/sync-sponsors.py used (developers.buymeacoffee.com/api/v1/supporters now redirects to a login flow and 502s). Sponsor data now comes from the buymeacoffee MCP server, which only works inside a live Claude conversation — there is no standalone script for the fetch step.

Step 1 — Ensure the MCP server is connected

Check with claude mcp list for a buymeacoffee entry. If missing, ask the user to run:

claude mcp add --transport http buymeacoffee https://mcp.buymeacoffee.com/connect

then complete the auth flow (e.g. via /mcp). Don't add the server yourself — it needs the user's own OAuth approval. Once connected, load its tools:

ToolSearch("select:mcp__buymeacoffee__get-recent-supports,mcp__buymeacoffee__get-recent-memberships")

Step 2 — Paginate every supporter source

There are four sources, each needing full pagination (limit=20, increment page until a page returns fewer than 20 results):

• get-recent-supports with support_type: "donation"
• get-recent-supports with support_type: "shop"
• get-recent-memberships with subscription_type: "membership"
• get-recent-memberships with subscription_type: "monthly_supporter"

This is a lot of tool calls with output you don't need to keep around — run it as a forked subagent (Agent tool, subagent_type: "fork") rather than inline, so the raw pagination noise doesn't fill the main conversation. Give the fork the exact cleaning rules from Step 3 so it applies them consistently and reports only the final list.

Step 3 — Clean and dedupe (mirrors v1's clean_name/escape_for_android)

• Skip entries where supporter_id is null (anonymous).
• Skip refunded payments (is_refunded: true) for donation/shop supports.
• Strip a leading URL prefix from the name (https://www.facebook.com/joe → joe).
• Strip a trailing email-domain suffix ([e-mail] → joe, tolerating spaces around @).
• Drop the literal name "Someone" (BMC's placeholder for a fully anonymous donor) — also watch for lookalike placeholder values (e.g. defaults.someone) that mean the same thing but won't match a literal string check; flag these to the user rather than silently keeping or dropping them.
• Dedupe case-insensitively across all four sources combined, keeping the first (i.e. most recent, since results are newest-first) occurrence.
• Escape for Android string resources: backslash-escape a leading @ or ?.

Step 4 — Flag anomalies before writing anything

Don't silently "fix" surprising data. In particular:

• A name containing a literal comma will visually blend into the comma-separated list — ask the user whether to keep it verbatim or sanitize it.
• Placeholder-looking names (see Step 3) — ask whether to drop or keep.

Use AskUserQuestion for each distinct anomaly found. Only proceed to Step 5 once resolved.

Step 5 — Write the final list and splice it in

Write the final, cleaned, one-name-per-line list to a scratch file, then:

python3 tools/sync-credits.py sponsors --sponsors-file /path/to/names.txt

This escapes each name and rewrites sponsors_text in app/src/main/res/values/contributors.xml, leaving contributors_text untouched.

When run as part of skymap.release

skymap.release Step 2 links here instead of v1's separate skymap.sponsors/skymap.contributors skills. Ask the user whether they want sponsors, contributors, or both refreshed for the release — don't assume.