peg-risk — Collateral & Listing Safety Gate
This skill answers one question before you ship: is this pegged asset safe to integrate right now, and what parameters should govern it? It is a pre-integration safety gate — not a monitoring dashboard. The output is a pasteable RISK-PARAMS block (asset class, warn/refuse/liquidate thresholds in bps, oracle staleness bound, direction sensitivity) you can drop directly into lending protocol config, listing logic, or routing guards. The methodology is derived from a public, MIT-licensed peg-risk framework (github.com/lrafasouza/pegana-replay) that has been running in production on Solana mainnet since 2026-05-29.
Why this is not the same as adjacent tools
| Other skills/tools | This skill |
|---|---|
| Pyth / Switchboard skills — give you a price | Judges whether that price has broken peg relative to intrinsic value |
| Token-intel / rug scanners — mint, honeypot, holder safety | Judges the asset's peg mechanism risk (market-vs-intrinsic spread, NAV/CR) |
| jupiter-lend / kamino — a borrower's LTV / liquidation health | Judges the collateral asset's own peg integrity before you accept it |
Operating Procedure
| What you need | Go to |
|---|---|
| Understand the 5 states and per-class thresholds | peg-states.md |
| Understand why naive depeg checks lie (and the 5 real guards) | failure-modes.md |
| Get intrinsic + market prices for a specific asset class | computing-spread.md |
| Run the full gate and produce a RISK-PARAMS verdict | assess.md |
| Links to APIs, methodology source, operationalization pointer | resources.md |
Progressive Disclosure
peg-states.md— 8 asset classes, 5 states, per-class threshold tables (bps + CR path), constants, and how to recalibrate defaults per asset.failure-modes.md— The 5 ways a naive depeg check lies, with the guard for each and the real mainnet incident that proved it.computing-spread.md— How to computeintrinsicandmarketfor each class using 2026 APIs (Pyth Hermes, Jupiter, Helius DAS, Sanctum, @solana/kit). Includes Tier 1 runnable adapters and Tier 2 on-chain-read recipes.assess.md— The gate workflow: classify asset → pull spread → run classifier → emit VERDICT + pasteable RISK-PARAMS block. Includes a worked jitoSOL example.resources.md— Canonical links to methodology source, Pyth, Helius, Jupiter, Sanctum, Switchboard, plus a 3-line "run this continuously" pointer.
Command & Agent
- Command:
/peg-assess <mint>— runs the gate end-to-end for a given mint address. - Agent:
peg-risk-analyst(model: opus) — performs the full classification workflow.
This skill is read-only. All on-chain interactions are reads only. The skill never signs transactions, never sends funds, and never requests a private key. Always use your own
HELIUS_API_KEY; the kit's bundled key returns 401. Thresholds in this skill are defaults derived from production data — treat them as starting points and recalibrate per asset.