Credit Compass

View saved version →

---
name: audit-loan-offer
description: Check a loan sanction letter, offer, quote or Key Facts Statement (KFS) for hidden costs and breaches of RBI rules, compare several offers by true cost, and check whether a loan app is legitimate. Use when the user uploads or pastes a loan offer or KFS, asks "is this a good offer", "is this loan app safe", or "which of these offers is cheaper".
---

# Audit a loan offer

Explicit instructions from the user override this workflow.

## One offer
1. Read the document the user shared. Extract: lender and lender type (bank / NBFC / HFC / digital app), loan type and purpose, sanctioned amount, rate and fixed/floating, tenure, EMI, APR, processing fee (and whether GST is included), other upfront charges, insurance premium and whether it is mandatory, prepayment/foreclosure charge, how late-payment penalties work, cooling-off period (digital loans), and where the money will be disbursed.
2. If a field is not in the document, leave it out — don't assume. Ask only about things that change the verdict: whether a fee was demanded before disbursal, and whether a KFS was given.
3. Call `audit_loan_offer`.
4. Present: the verdict and headline first; then a table of findings ordered violation → warning → ok, each with the RBI source link; then the questions to ask the lender (as a ready-to-send message).
5. If the verdict is `walk-away` because of an upfront payment demand or third-party disbursal, say plainly not to pay, and give the reporting routes from the result.

## Several offers
Extract each offer the same way, then call `compare_loan_offers`. Lead with the cheapest offer by total cost and say explicitly when a lower headline rate is not the cheaper loan. If tenures differ, say the totals aren't like-for-like and compare APR.

## Loan apps
For a digital loan app, call `check_loan_app_safety`. Ask the unanswered questions in one message (yes / no / unsure). If the verdict is `stop`, tell the user not to share contacts/photos access or pay anything, and link the RBI directory from the result.

## Don't
- Don't call a lender's conduct illegal unless the finding is a `violation` with a cited rule; otherwise say "worth questioning".
- Don't repeat account numbers, PAN, Aadhaar or addresses from the document back to the user.

View saved version →

---
name: borrowing-readiness
description: Assess whether someone is ready to take a loan and how lenders will see them — reading their credit report, overall financial health, irregular freelance or gig income, and what-if scenarios like taking a new loan or closing one. Use for "why was my loan rejected", "explain my CIBIL report", "am I ready for a loan", "I'm a freelancer, how much can I borrow", "what happens if I take a ₹5L personal loan".
---

# Borrowing readiness

Explicit instructions from the user override this workflow.

## Credit report
If the user shares a CIBIL / bureau report, read the score, total card limits and balances, hard enquiries in the last 6 months, 30+ day late payments in 24 months, any settled / written-off / SMA status, oldest account age, and active secured vs unsecured loans. Call `explain_credit_report`. Explain each factor plainly and give the time-based habits from the result. Be explicit that accurate entries cannot be removed; for inaccurate ones, explain the free bureau dispute route from `disputeRights`. Never suggest paid "credit repair".

## Financial health and what-ifs
Collect take-home income, fixed monthly expenses, each existing loan (outstanding, EMI, rate, months left), emergency savings and other savings. Call `check_financial_health`. For a "what if" question, call it once with `whatIf` and present before → after.

## Irregular income
For freelancers, gig workers and business owners, get 6–12 months of income (or total the credits in a bank statement the user shares, excluding own-account transfers, refunds and loan disbursals). Call `assess_variable_income`. Explain that banks average two years of ITR income and plan EMIs around the lowest months.

## Answer format
Start with a one-line readiness verdict, then the 2–3 factors that matter most, then concrete next steps with timeframes. Don't repeat account numbers or personal identifiers from documents.

View saved version →

---
name: home-loan-decision
description: Help an Indian home buyer decide whether they can afford a home loan, what rate they will likely get, which bank and government schemes fit, and what the loan will really cost. Use when someone is planning to buy a flat or house, asks "can I afford a ₹X home loan", "how much home loan can I get", or "which bank for my home loan".
---

# Home loan decision

Explicit instructions from the user override this workflow. Skip steps they don't want.

## Answer first, then refine
Start computing as soon as you know the **loan amount** (or property price) and **monthly take-home**. Do not open with a list of questions. Fill every gap with a stated assumption:
- Tenure 20 years · existing EMIs ₹0 · living expenses ≈ 35% of take-home
- Rate: the median from `get_current_interest_rates`
- Salaried private-sector employee · CIBIL 750 · ready property

Say the assumptions in one line. The cards have sliders, so the user can adjust amount, rate and tenure themselves. After the result, ask at most **one** question — the one that would change the verdict most (usually existing EMIs or expenses).

Never ask for PAN, Aadhaar, account numbers or documents with them.

## Steps
1. `get_current_interest_rates` (loanType `home`) — never guess a rate.
2. `check_loan_affordability` with the user's numbers (and the assumptions above) at the median advertised rate. If the verdict is `tight` or `unaffordable`, lead with that and the recommended loan amount before anything else.
3. `predict_my_interest_rate` — the rate this borrower will likely get, and the bank that fits best.
4. `calculate_loan_cost` at the predicted rate — EMI, total interest, APR, tax benefit, saving habits.
5. `find_government_schemes` (purpose Home). For state schemes, check the user's state first.
6. `list_loan_red_flags` (Home, severity high) — pick the 3 most relevant to their situation (e.g. under-construction → pre-EMI trap).

## Answer format
1. **Verdict in one line** — affordable / stretch / not yet, with the safe loan amount.
2. A short table: EMI, FOIR, likely rate, total interest, APR.
3. Best-fit banks (max 3) and why.
4. Schemes that match, with the official portal link.
5. Three things to check before signing.
6. One line: these are estimates, not a loan offer; link to the relevant `openOnSite` page.

## Don't
- Don't present advertised minimum rates as what the user will get.
- Don't recommend a specific lender as "the best"; say "best fit for this profile" and show the alternatives.

View saved version →

---
name: manage-existing-loans
description: Help someone with loans they already have — what an RBI repo rate change does to their EMI, whether to transfer a loan to a cheaper lender or ask their bank to cut the rate, whether to prepay, and how to clear several loans or credit card balances faster. Use for "RBI hiked rates, what happens to my EMI", "should I do a balance transfer", "write to my bank to reduce my rate", "prepay or invest", "how do I get out of these loans".
---

# Manage existing loans

Explicit instructions from the user override this workflow.

## Rate change ("RBI changed the repo rate")
1. `get_rbi_policy_rates` to state the current repo rate and the next MPC date.
2. Ask the loan's benchmark if unknown (most bank home loans since Oct 2019 are repo-linked / EBLR; older ones MCLR).
3. `calculate_repo_rate_impact` with the user's outstanding, rate, remaining months and the change in bps. Show both options — higher EMI vs longer tenure — and note banks extend tenure by default; RBI lets the borrower choose.

## Cheaper rate available
1. `get_current_interest_rates` for the loan type.
2. `build_rate_negotiation_brief` — then draft a short, polite letter or email to the current lender using its facts (gap, savings, leverage, the specific ask). Asking the current lender to reprice comes first.
3. If the user has a competing offer, `analyze_refinancing` with that offer's rate and fees and give the verdict (switch / negotiate first / stay).

## Prepay or not
`simulate_prepayment` for the user's amount. Compare the interest saved (the "return") with what the money would earn safely, and mention keeping an emergency fund first. If they have several loans, prepay the highest-rate one.

## Several loans or card balances
`plan_debt_payoff` with each loan and any extra monthly amount. Recommend avalanche (highest rate first) unless the user wants quick wins (snowball). If they ask about consolidation, include `consolidationRate` and compare total interest. This is a self-managed repayment plan — never suggest settlement, skipping payments or "debt relief" services, which damage the credit report.

## Answer format
Lead with the number that answers the question, then a small table, then one concrete next step. Close with a one-line estimate disclaimer.

View saved version →

---
name: study-abroad-funding
description: Plan how an Indian student funds a degree abroad (USA, UK, Canada, Australia, Germany) — loan amount at live exchange rates, interest during the moratorium, EMI after the course, visa proof-of-funds rules, lenders with and without collateral, documents and scams to avoid. Use for "education loan for MS in the US", "how much loan for UK masters", "GIC / blocked account / I-20 funds".
---

# Study-abroad funding

Explicit instructions from the user override this workflow.

## Collect
Country, course length, total tuition + living cost (in the country's currency — if the user only has tuition, ask for living cost or use the cost table from the result and say so), family contribution, scholarship, and whether property collateral is available.

## Steps
1. `plan_study_abroad_loan` with the figures.
2. `get_current_interest_rates` (loanType `education`) if the user wants lender rates beyond the country lender list.
3. If they ask about repayment after graduating, use the EMI from the result; if they have an expected salary, compare the EMI with 30–40% of take-home pay.
4. Use `find_government_schemes` (purpose Education) for interest subsidy schemes when family income is under ₹12L.

## Answer format
1. Loan needed and the EMI after the moratorium (one line each).
2. The visa funds rule for that country — what document, how much, by when.
3. Lenders: up to 4, split collateral / no collateral, with limits.
4. Moratorium tip: paying simple interest during the course reduces the EMI.
5. Country red flags from the result (max 3) and the documents checklist link.

Mention that exchange rates move and the result's rate is today's.