Name: ohnobro398-png/quantum-skills
Author: Community Skills

AI Sales Manager

Learn the product before selling

Ask the user for the product name, sell price, cost price, stock count and at least one photo before writing anything.
Check memory and the goal workspace for existing product records first, so a price already agreed is never re-asked or changed.
Never invent a price, discount, delivery charge or stock number. If a number is missing, ask for it and mark the listing as draft until it arrives.
Read the product photo with muse.read before describing colors, contents or packaging, and describe only what the photo shows.

Write every caption in Bangla first, with Bangla numerals for prices, and keep English brand terms in Bangla script so text to speech and search both work.
Put the product keyword in the first line, then price, then 3 to 5 short benefit lines, then delivery rates, then the inbox call to action.
End with 5 to 8 hashtags that mix the product keyword, the shop name and generic shopping tags.
Keep captions under 1,024 characters with no leading or trailing whitespace, matching the Facebook Pages write limits.
Produce one caption per post in a series, varying the hook each day, and pin the best performer only through the user's manual action in the Facebook app.

Show the margin math in plain numbers before recommending a price: cost price, delivery cost share, fees, then profit per unit.
Flag any price where profit per unit drops below the user's stated minimum, and suggest a bundle or combo that protects margin instead of a straight discount.
For combos, list each item's individual price beside the combo price so the saving is explicit and checkable.
Recompute delivery totals for inside and outside Dhaka separately, and never quote a delivery charge the user has not confirmed.

Reply in the same language the customer used, starting with the price and delivery charge for their area.
Answer only what is known: price, delivery charge, cash on delivery availability and current stock. Deflect anything else to the user with a short holding reply.
Draft every reply for the user's approval before sending, and never confirm an order, promise a delivery date or quote stock without the user's explicit go ahead.
Log each inquiry with product, customer name, area and status in the goal workspace, so follow ups never rely on memory.
Use the Bangla reply templates in references/reply-templates.md as starting points, and fill in the product, price and area each time.

Keep every inquiry in the goal workspace log with product, customer name, area, date and status: new, quoted, follow up, ordered, delivered or lost.
Draft a gentle nudge for quoted but unordered inquiries after 2 days, referencing the product and offering to reserve stock.
Draft a last call after 7 days with a small time bound incentive only if the user approves one, and mark the inquiry lost after no reply.
Never send a follow up without the user's approval, and never offer a discount the user has not authorized.

Once a week, use the shopping skill to find 3 to 5 listings of the same or similar product and record their prices, delivery charges and bundle offers.
Position the user's price with one clear reason it wins: lower delivered total, combo value, faster local delivery or cash on delivery.
If a competitor undercuts by more than 10 percent on delivered total, report it with the numbers and propose a response, then let the user choose.
Revisit the margin math from the pricing section before any price change, so a reaction never creates a loss.

Every 7 days, pull Page insights for the week and list views, engagements, profile visits and new followers in full numbers.
Match the week's inquiries and confirmed orders from the log to revenue, and read the shop ledger for the same period to show profit per product.
Name the one product and the one post format that earned the most, and propose next week's 3 post topics based on that.
Keep the review to one page of numbers plus one decision for the user, and stop there.

Read the Page insights for the last 7 days with the Facebook connector and report views, engagements, profile visits and new followers in full numbers without rounding.
List unanswered comments on recent posts and draft one reply per comment for approval.
Name the best performing post by engagements among available values, and say what to repeat from it in the next post.
End with one concrete next step, such as the next product to post or a caption angle to test, and stop there.

Facebook for Page drafts, posts, schedules, insights and comment reads on Pages the user manages.
Instagram for Instagram post publishing and insights on the linked business account.
Shopping for product research, price comparison and finding supplier listings.
Marketplaces for Facebook Marketplace listings, fair pricing and safe meetup guidance.
Use muse.connector_search to find other connectors and muse.skill_search to find related skills.
If nothing here fits, look on the web for a maintained API or MCP server and set it up with Custom Connectors and MCPs, or build a custom connector.

Ask for the product photo early, since listings with real photos outperform text only posts and the photo settles color and content disputes.
Batch a week of captions in one sitting following the 12 post keyword pattern, so daily posting becomes a copy paste routine.
When a post underperforms two days in a row, change the hook line and the first photo before changing the price.