Hyperliquid review

Reconstruct what happened from exchange records, then grade process separately from outcome.

If a skill rule blocks progress, cite its file and exact rule, explain the missing input or authority, and continue independent work within this skill's boundary. User instructions govern workflow and style defaults; they do not bypass tool or financial controls.

Tool selection

When connecting or choosing tools, read official tools. Reuse a suitable maintained upstream capability before implementing new integration code; tool adoption stays optional and never supplies wallet authority.

First task and connected workflow

Reconstruct one completed trade from fills, fees and funding; use the bundled arithmetic script on explicit inputs and mark any missing exchange evidence. Read agent integration and handoffs on first use or when a tool or related skill is missing.

Task handling

Use the existing ticket and conversation to scope the review. Ask only for missing identity or history that affects the result; complete supported accounting while marking gaps unknown. Lead with the process finding and evidence. A follow-up question does not reset the review or authorize an exchange action.

Keep the user's strategy ID/revision and separate historical simulation, forward paper records and verified live fills in summaries. Do not mix their performance series. Reuse the requested base currency and review window, explaining valuation sources. If the original risk budget is unknown, report R as unknown while completing supported cash accounting; a zero gross PnL denominator makes cost ratios undefined, not zero.

Workflow

1. Bind the network, user account, market and DEX, review window, ticket ID, account mode, and known client order IDs.
2. Gather review evidence: ticket, approvals, order status, historical and detailed open orders, fills, funding, ledger changes, positions, portfolio samples, and incident records.
3. Deduplicate fills by trade ID. Preserve maker or taker status, fee token, signed funding, partial fills, and position before each fill.
4. Compute gross PnL, fees, funding, net PnL, slippage against the ticket, holding time, protection gaps, and result in the ticket's original risk unit. Use node scripts/review.mjs --help for simple round-trip arithmetic.
5. Grade process and incidents before reading the profit or loss verdict.
6. Append facts to the journal contract. Corrections are new entries, never rewrites.
7. Deliver one repeatable finding with an owner or next test. Do not produce a list nobody will act on.

Evidence rules

• Chat shows intent. Orders and fills show exchange activity. Current account state shows what remains.
• Portfolio graphs are sampled and are not precise trade accounting.
• A missing client order ID weakens reconciliation and is itself a process finding.
• Funding is signed from the account's perspective. Do not reverse it twice.
• Mark unrealized PnL separately from realized PnL.
• Unknown or unavailable history stays unknown. Do not fill gaps from screenshots or memory.

Process grade

Use clean, minor break, or major break. Major breaks include an unapproved write, material ticket drift, duplicate send on an unknown result, exposure above limits, or an unprotected position. A profitable major break remains a major break.

Boundary

Do not recommend a new trade, modify controls to excuse an outcome, or perform any exchange write. Route new research to hyperliquid-analyze and any new action through a fresh plan.