abhiroopb/saving-cash-rounding-feedback

Saving Cash Rounding Feedback

Appends seller feedback to the Feature Feedback tracking table in Google Docs.

STOP if the gdrive skill is not available or the tracking doc cannot be accessed. Also stop if the feedback text or seller identity (email or merchant token) is missing — ask the user to provide them.

Target Document

• Doc ID: 1WAPVMavQbnmnU83AWY7T1vGOCo92oZqSbhBMI6ZPhHE
• Tab ID: t.yg7h7jh60g59 (this is the "Post-Launch" child tab under "Seller feedback")
• URL: https://docs.google.com/document/d/YOUR_DOC_ID/edit
• Important: When using docs get, you MUST use --include-tabs to see child/sub-tabs. The Post-Launch tab is a child of the "Seller feedback" tab and won't appear in the top-level tab list. Use find_tab recursively through childTabs.

Workflow

When the user says to save feature feedback, follow these steps:

Step 1: Identify the feedback source

The feedback can come from two sources:

Option A: Email

The user will provide (or you'll extract from the current context):

• The feedback text (verbatim, word for word — do NOT modify it)
• The seller's email address

If either is missing, ask the user.

Option B: Slack link

The user will provide a Slack message URL (e.g., https://your-org.slack.com/archives/CXXXXXX/pXXXXXXXXXX).

1. Load the slack skill and read the message using get-channel-messages with the channel ID and thread timestamp extracted from the URL. Use --full-text to avoid truncation.
2. Extract the feedback text verbatim from the message.
3. If a merchant token (e.g., MT: XXXXX or similar) is mentioned in the message, extract it directly — skip the Regulator lookup in Step 2.
4. The Source column should be the raw Slack link (not "Email from…").

Step 2: Look up merchant token from Regulator

Skip this step if the merchant token was already extracted from the Slack message.

Search Regulator for the seller using their email address.

Method 1: GraphQL omniSearch

sq curl -s -X POST 'https://graphql-gateway.internal.example.com/graphql' \
  -H 'Content-Type: application/json' \
  -H 'apollographql-client-name: amp-regulator-skill' \
  -H 'apollographql-client-version: 1.0.0' \
  -d '{"query": "query { omniSearch(query: \"<EMAIL_ADDRESS>\") { merchantToken } }"}'

Method 2: Regulator Advanced Search UI (fallback)

If the API returns null, provide the user with the Regulator advanced search link so they can look it up manually:

https://regulator.internal.example.com/o/advanced-search?searchParams=%7B%22email%22%3A%22<URL_ENCODED_EMAIL>%22%7D&searchTarget=email

Note: The advanced search is a client-side UI and cannot be queried via API. Always try the API first, then provide the link.

Extract the merchantToken from the result. If no merchant is found, note "Not found" in the merchant token column, provide the Regulator search link, and inform the user.

Step 3: Read the current document structure

Get the document structure to find the table and its indices:

cd ~/.agents/skills/gdrive && uv run gdrive-cli.py docs get 1WAPVMavQbnmnU83AWY7T1vGOCo92oZqSbhBMI6ZPhHE

Find the table in the tab t.yg7h7jh60g59. Identify:

• The table's element index
• The last row index (to insert after it)
• The cell start indices for each column

Step 4: Add a new row to the table

Insert a new row at the bottom of the table:

cd ~/.agents/skills/gdrive && uv run gdrive-cli.py docs batch-update 1WAPVMavQbnmnU83AWY7T1vGOCo92oZqSbhBMI6ZPhHE \
  --requests '[{"insertTableRow": {"tableCellLocation": {"tableStartLocation": {"index": <TABLE_INDEX>}, "rowIndex": <LAST_ROW_INDEX>, "columnIndex": 0}, "insertBelow": true}}]'

Step 5: Populate the new row cells

After inserting the row, re-read the doc structure to get updated indices, then populate the three columns:

1. Column 1 — Feedback: Copy/paste the feedback text verbatim (do not change anything)
2. Column 2 — Merchant Token: The merchant token from Regulator lookup
3. Column 3 — Source: "Email from <email_address>" (for emails) or the raw Slack link (for Slack messages)

Use batch-update with insertText requests to populate each cell:

cd ~/.agents/skills/gdrive && uv run gdrive-cli.py docs batch-update 1WAPVMavQbnmnU83AWY7T1vGOCo92oZqSbhBMI6ZPhHE \
  --requests '[
    {"insertText": {"location": {"index": <COL1_INDEX>}, "text": "<FEEDBACK_TEXT>"}},
    {"insertText": {"location": {"index": <COL2_INDEX>}, "text": "<MERCHANT_TOKEN>"}},
    {"insertText": {"location": {"index": <COL3_INDEX>}, "text": "<SOURCE>"}}
  ]'

Important: When inserting text into multiple cells in a single batch, process the cells in reverse index order (highest index first) so earlier insertions don't shift the indices of later ones.

Step 6: Confirm

Tell the user:

✅ Saved feedback to the Feature Feedback tracking doc.

• Merchant token: <token>
• Source: <source (email or Slack link)>
• Doc: [link to doc]

Notes

• Always copy the feedback word for word — never summarize, clean up, or rephrase.
• If the Regulator lookup returns multiple merchants, list them and ask the user to confirm which one.
• The tab ID t.yg7h7jh60g59 is important — make sure you're looking at the right tab when reading the doc structure.