> For the complete documentation index, see [llms.txt](https://forge-labs.gitbook.io/forge-docs/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://forge-labs.gitbook.io/forge-docs/hallmark/methodology.md).

# Methodology

Hallmark is the underwriting framework behind every ForgeYields allocation. It scores risk in **three layers** — each a self-contained rubric — that compose into a single Global Risk Score per strategy.

<figure><img src="/files/QZiYJ0YnjgHVOQDwXWBb" alt="Hallmark methodology — Protocol Risk × 0.35 + Asset Risk × 0.25 + Strategy-Specific Risk × 0.40 = Global Risk Score, with eligibility cutoff at 7.5 and hard-cap overrides on single-criterion thresholds"><figcaption><p>How the three layers compose into a Global Risk Score. Hard-cap thresholds on any single criterion can exclude a strategy regardless of composite GRS.</p></figcaption></figure>

## The three layers

| Layer                  | Scope                                                   | Output                     |
| ---------------------- | ------------------------------------------------------- | -------------------------- |
| **Layer 1 — Protocol** | Per-protocol assessment (Aave, Morpho, Curve, Pendle…)  | Protocol Risk Score (1–10) |
| **Layer 2 — Asset**    | Per-asset assessment (USDC, wstETH, sUSDe, PT-tokens…)  | Asset Risk Score (1–10)    |
| **Layer 3 — Strategy** | Per-strategy composite (the actual deployable position) | Global Risk Score (1–10)   |

**Scoring convention:** 1 = lowest risk, 10 = highest risk. **Eligibility cutoff: GRS ≤ 7.5.**

***

## Layer 1 — Protocol Risk

Six weighted criteria assess the systemic risk of each protocol ForgeYields interacts with.

| #      | Criterion                          | Weight | What it measures                                                                     |
| ------ | ---------------------------------- | ------ | ------------------------------------------------------------------------------------ |
| **C1** | Audit Status                       | 25%    | Audit count, auditor tier, recency, unaudited code delta                             |
| **C2** | TVL History                        | 10%    | Absolute TVL, 30d drawdown, ATH distance                                             |
| **C3** | Governance Quality                 | 20%    | Multisig configuration, timelock duration, upgrade controls, custody mode            |
| **C4** | Incident History & Operational Age | 20%    | Past exploits, loss magnitude, response, operational track record                    |
| **C5** | Smart Contract Risk                | 15%    | Code complexity, external dependencies, upgradeability pattern, off-chain dependence |
| **C6** | Team & Transparency                | 10%    | Doxxing status, public reputation, communication consistency                         |

**Formula:** `PR = 0.25·C1 + 0.10·C2 + 0.20·C3 + 0.20·C4 + 0.15·C5 + 0.10·C6`

### Notable refinements

* **Type A vs Type B contracts (C1):** Immutable-core protocols are not penalized for audit age — only audit quality and unaudited code delta matter. Upgradeable-core protocols face the full recency rubric.
* **Custody Mode Recognition (C3):** MPC threshold-signature setups are treated distinctly from plain multisigs and EOAs, reflecting their operational security profile.
* **Off-chain dependence (C5):** Protocols relying on CEX accounts for delta-hedging or cloud services for core operations score a minimum of 7 — these counterparty/operational risks are not mitigated by smart-contract audits.

### Chain Risk (CRS)

Chain-level risk (Ethereum, Starknet, Monad, etc.) is scored separately from protocol risk using the **N-criteria rubric** and published as `Chain Risk Score (CRS)`. Per-chain protocol re-deployments inherit chain risk but are not treated as new protocols.

***

## Layer 2 — Asset Risk

Each asset (stablecoin, LST, LRT, wrapped asset) is scored on backing, peg behavior, and redemption mechanics.

| #      | Criterion          | What it measures                                                                   |
| ------ | ------------------ | ---------------------------------------------------------------------------------- |
| **A1** | Backing Quality    | Collateral type, attestation cadence, transparency                                 |
| **A2** | Peg History        | Historical deviation magnitude and recovery time                                   |
| **A3** | Liquidity & Exit   | Onchain liquidity depth, async exit credit (queue/cooldown), redemption guarantees |
| **A4** | Collateral Backing | Bridge risk, custody assumptions, dependency on canonical bridges                  |
| **A5** | Issuer/Operator    | Counterparty risk of the issuer or operator                                        |

**Async exit credit (v3.7):** Assets with built-in async redemption queues receive partial credit on A3 even when onchain liquidity is thin, provided the queue is honored and bounded.

***

## Layer 3 — Strategy Risk

Layer 3 combines protocol and asset risk with strategy-specific factors to produce the **Global Risk Score (GRS)** used for eligibility.

**Formula:** `GRS = PR × 0.35 + AR × 0.25 + SSR × 0.40`

Where **SSR (Strategy-Specific Risk)** is computed from rubric-specific criteria depending on strategy type.

### Strategy types

| Type        | Description                                                              | Rubric |
| ----------- | ------------------------------------------------------------------------ | ------ |
| **Type 1**  | Looping / leveraged lending (e.g. wstETH/WETH looper on Aave)            | S1–S5  |
| **Type 2A** | Pendle LP (PT/YT split)                                                  | S1–S5  |
| **Type 2B** | Curve/Balancer LP                                                        | S1–S5  |
| **Type 2C** | Concentrated liquidity LP (Uniswap v3, Ekubo)                            | S1–S5  |
| **Type 3**  | Direct lending / non-looped supply                                       | S1–S5  |
| **Type W**  | **Wrapper Vault** (Ipor Fusion, MetaMorpho, Yearn V3, ERC-4626 wrappers) | X1–X5  |

### Type W — Wrapper Vault (v4.1)

Type W exists because depositing into a third-party permissioned vault introduces a **two-layer trust model** that Types 1/2/3 don't cleanly capture. When ForgeYields directly executes a strategy, all risk lives at L1 + L2 + strategy mechanics. When ForgeYields deposits into a vault that internally executes a strategy, the **curator/atomist, exit queue, fee structure, and vault maturity** become first-order concerns.

<figure><img src="/files/Fn8Sfa3EKInqGFWnlzEW" alt="Type W trust model — Depositor → ForgeYields fyToken → Wrapper Vault (curator/atomist) → Underlying Strategy. X-criteria measure trust at the wrapper layer, with X2 (Curator/Atomist Trust) carrying 30% weight as the defining differentiator."><figcaption><p>Type W adds a 4th party (the Wrapper Vault) to the capital flow. The X-criteria measure trust at that wrapper layer — X2 (Curator/Atomist Trust) at 30% weight is the defining differentiator.</p></figcaption></figure>

| Criterion                       | Weight | What it measures                                                                               |
| ------------------------------- | ------ | ---------------------------------------------------------------------------------------------- |
| **X1** Underlying Strategy Risk | 25%    | What the wrapper does economically (looping, LP, lending) — scored as if direct                |
| **X2** Curator/Atomist Trust    | 30%    | Who can move funds. EOA = 9, MPC-attested = 6–8, multisig + timelock = 3–5                     |
| **X3** Exit Mechanism           | 20%    | Atomic = 1–2, queue ≤ 1d = 3–4, queue ≤ 7d = 5–6, queue > 7d or pause history = 7–9            |
| **X4** Fee Structure            | 15%    | Transparent low = 1–3, standard = 4–5, opaque/high = 6–8, unilateral change = 8–9              |
| **X5** Vault Maturity           | 10%    | >2y + >$100M = 1–3, >1y + >$10M = 4–5, <1y or <$10M = 6–8, <3 months or recent incident = 8–10 |

### Recursive Strategy Collateral Rule (v3.8)

If a strategy uses another scored strategy or vault token as collateral, the dependency cascades: any score change in the underlying triggers a re-score of the dependent. Hallmark's validators enforce this graph integrity.

***

## Versioning

The methodology is versioned (currently v4.2). Each amendment is a published, dated document with backwards-compatibility notes. All scores reference the methodology version they were computed under, so historical scores remain interpretable even after the rubric evolves.

| Version | Date       | Headline change                                                                                       |
| ------- | ---------- | ----------------------------------------------------------------------------------------------------- |
| v3.5    | 2026-04-23 | C4 max rule                                                                                           |
| v3.6    | 2026-04-23 | C5 cross-chain, C1 audit coverage                                                                     |
| v3.7    | 2026-05-06 | L2 A3 async exit credit + WATCHLIST band 6.0–6.5                                                      |
| v3.8    | 2026-05-07 | L3 Recursive Strategy Collateral Rule                                                                 |
| v3.9    | 2026-05-07 | L1 C3 Custody Mode Recognition (MPC)                                                                  |
| v4.0    | 2026-05-12 | Chain Risk (CRS) as a distinct layer                                                                  |
| v4.1    | 2026-05-18 | Type W (Wrapper Vault) for Ipor Fusion / MetaMorpho-style vaults                                      |
| v4.2    | 2026-05-22 | L3 procedural: GRS always computed; verdict derived separately (cascade exclusion ≠ skip computation) |