feat(risk): freeze TQQQ ETF-only research mandate

[Pigbibi]

Scope

• freeze tqqq_etf_only_research_v1 as an exact RESEARCH_ONLY ETF-only mandate
• bind candidate/state/stop/breaker/drawdown evidence into risk-gate assessment v2
• enforce TQQQ/BOXX caps, one-nonzero rule, no margin stacking, cash-only over-cap normalization, and zero execution authority
• extend pure sizing for exact TQQQ/BOXX product identities

Safety boundaries

• no provider or real backtest calls
• no platform/runtime/config changes
• no paper/shadow/live/order/capital authority
• target containment remains a separate fail-closed prerequisite

Evidence

• tests-first RED captured on exact base 190b7577e3d4cd826f053e4b899c13e127191097
• focused: 9 passed, 43 subtests passed (later expanded coverage)
• full: 932 passed, 1 skipped, 26 warnings, 131 subtests passed
• ruff check ., compile, package, diff, scope, and secret gates passed

Bound freeze SHA-256: 4e8896cdd16eefff7631ce8c491cdba283e3dca19770735fd7a7f0b53fd498ce
Containment SHA-256: 4aeacd014221cfc8be6541f4d083463c4e6b4f12df48d50d07157d30f2db84dc

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: d687678322

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".

[chatgpt-codex-connector]

Enforce the budget against completed-session equity

For this mandate, loss_budget_equity_reference is fixed to completed_session_equity, but this comparison uses a weight derived from portfolio_snapshot.total_equity (including when a target_value is normalized) and the gate never receives the completed-session equity value. When current equity differs from the completed-session reference, the one-percent budget is therefore applied to the wrong base; for example, with completed-session equity of $100k and current equity of $120k, a $24k BOXX target becomes weight 0.20 and passes this check while its modeled 5% loss is $1,200, or 1.2% of the mandated reference. Bind the completed-session equity into the evidence and compare the modeled dollar loss against that value.

Useful? React with 👍 / 👎.

[chatgpt-codex-connector]

Check stop readiness against current TQQQ holdings

This derives tqqq_active only from the proposed decision, not from the portfolio snapshot. If the account currently holds TQQQ but the new decision targets cash (or otherwise omits TQQQ), the gate accepts stop_intent_ready=False and missing fill identities even though the position remains exposed until the exit fills; a snapshot with current TQQQ exposure and an empty decision can consequently receive APPROVE. Determine active TQQQ exposure from the current portfolio state, or require the stop evidence until the holding is confirmed flat.

Useful? React with 👍 / 👎.