agentsclimarketplace

Raroc repricing

Skill alzadjaliaafra-hash/murshidi-knowledge-layer/models/raroc-repricing

Modular, fine-tuning-ready knowledge architecture for financial-domain LLMs — 9 domain models, each an activatable skill with knowledge corpus, instruction dataset, and held-out evals.

Install
npx -y skills add alzadjaliaafra-hash/murshidi-knowledge-layer --skill raroc-repricing

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

2 things to look at

  • 27 days oldThe repository was created 27 days ago. New is not bad, but a brand new repository carrying a familiar-sounding name is the shape a typosquat arrives in, and there has been no time for anyone else to find a problem with it.
  • 0 stars0 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

What its author says it does

Copied from the file, not written here

Sector-adjusted RAROC repricing methodology — a three-layer architecture that augments the classic risk-adjusted return on capital hurdle with a sector-specific minimum pricing floor and a borrower-WACC advisory signal, identifying underpriced credit facilities. Activate with /raroc, "repricing", "RAROC", "pricing floor", or "underpriced facility".

SKILL.md

3.8 KB, 811 tokens by cl100k_base, as published. Nobody here has run it

Sector-Adjusted RAROC Repricing (RAROC 2.0)

An enhanced corporate-banking repricing methodology that adds a sector-adjusted cost-of-capital layer to the standard RAROC framework. It identifies underpriced facilities using bottom-up unlevered sector betas and generates actionable repricing signals for relationship managers.

Three-Layer Architecture

LayerNameFormulaOutput
1RAROC Hurdle[NII + Fees + Funding Benefit − ECL − OpEx] ÷ [RWA × k]Hard gate at the institutional hurdle rate (e.g., 12.5%). Unchanged from the legacy model.
2Sector Minimum RateMin_Rate = COF + LP + (β_u × ERP × E_wt) + ECL + OpEx + CapPricing floor per sector — the relationship manager's repricing target, in bps gap.
3Excess Return SignalExcess = Lending Rate − WACC_borrowerAdvisory signal in bps, computable daily. Detects subsidy to borrower equity.

The combined signal — GREEN / AMBER / YELLOW / RED — reads all three layers simultaneously. A facility can pass Layer 1 while flashing RED on Layers 2 and 3: that is the structural blind spot the methodology is designed to expose. Layer 1 is untouched; Layers 2 and 3 are pure additions, so the model deploys without disturbing existing approval gates.

Beta Computation

β_u = Σ [ β_L_i ÷ (1 + (1 − t_i) × D/E_i) ] × w_i

Sector unlevered betas are built bottom-up from emerging-market comparables (Damodaran methodology), value-weighted across the peer set, then applied through the equity-weight term E_wt in the Layer 2 floor.

Governance Lessons (committee-validated)

The methodology survived a full credit-committee sovereign probe. Three amendments emerged that now bind every application of the model:

  1. Turnaround timelines must be realistic. A breakeven claim was corrected from one year to a 24–36 month trajectory once the required revenue growth (+181%) was made explicit. Always state the growth rate a breakeven claim implies.
  2. Revaluation reserves are not lendable collateral. Forced-sale discounts of 30–50% strip "hidden structural value" arguments. Treat revaluation reserves as neutral.
  3. A market price discount is not automatically a signal. Illiquidity, non-distributable reserves, and competitive import pressure can fully explain a discount. Default interpretation: neutral until proven asymmetric.

Two pre-conditions gate every repricing memo: a group-level RAROC pull across all related counterparty accounts, and a sector concentration check against the existing portfolio.

Application Workflow (8 steps)

  1. Pull facility economics (NII, fees, funding benefit, ECL, OpEx, RWA)
  2. Compute Layer 1 RAROC against the hurdle
  3. Map the borrower to its sector beta cohort and compute β_u
  4. Build the Layer 2 sector minimum rate; record the bps gap to the current rate
  5. Estimate borrower WACC and compute the Layer 3 excess return signal
  6. Combine into the four-colour signal
  7. Run the group-level and concentration pre-condition checks
  8. Issue the repricing memo with all three layers, the signal, and the target rate shown

Resources

  • knowledge/methodology.md — full equation register and worked examples (anonymized)
  • dataset/train.jsonl — instruction-tuning pairs
  • eval/eval.jsonl — held-out evaluation questions

Keep looking

Skills are one crate of 328,083. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.