Feature plan: Step 1 · Step 2 · Step 3 · Step 4

Debug mode: enabled Effort study: railspike-effort / v3/high_traffic_toggle; ledger. Standing P2 enrollment. Extended estimates: enabled. Approval: Step 2 (scope v3, Option A) approved; Step 3 drafted.

Cycle estimate: 0.20

Completion Contract

• Normal entry: Tools → Feature Flags as a flag manager; one high-traffic row directly beneath the heading, matching flag rows, with the shared switch and on/off text.

• End-to-end outcome: signed enable and disable through the existing prepare/finalize flow; switch matches ./v3 high-traffic status.

• Required recovery: short error with reload and ./v3 high-traffic restore; switch stays retryable, including by keyboard.

• Deployment/external: none; no production mode change, Apache or schema change.

• Release condition: both stages verified in a real browser, toolbar behaviour unchanged, non-managers see no row, phone/desktop, light/dark and Word97/Chicago/Midnight checked.

Key Risks

• Flag-script coupling: reusing data-feature-flag-* attributes would put the row into filters/counts/saving. Validate in Stage 1 by asserting toolbar counts; mitigation: layout classes only, own data-high-traffic-* hooks.

• Theme rendering: the switch outside a flag group may misalign in icon themes. Validate in Stage 2 screenshots; mitigation: reuse existing row/switch classes, minimal CSS.

• Browser test drift: tests/browser/high_traffic_control.mjs keys on current markup and "CLI recovery" text. Update in Stage 1 with stable hooks.

Stage 1

Estimate: 0.10

• Goal: replace the card with a flag-style row and drive the shared switch from the existing control script.

• Dependencies: approved Step 2.

• Expected changes:

  • templates/pages/feature_flags.php: remove the card; render one row beneath the h1 (label, staleness line, .switch button with role="switch"/aria-checked, on/off state text, hidden status), manager-only as now.

  • public/assets/high_traffic_control.js: update aria-checked and state text instead of button text; shorter failure message keeping reload and CLI restore.

  • tool-details.css: only if needed to space the standalone row; no new switch style.

  • tests/browser/high_traffic_control.mjs: target the new hooks; assert toolbar "N of N flags shown" excludes the row.

• Verification approach: browser test (signed on/off, failure, keyboard retry); php tests/run.php HighTrafficTest and existing Feature Flags tests; PHP lint.

• Estimate details: driver: markup and script adjustments plus test update; uncertainty: browser test harness setup time.

• Risks or open questions:

  • Impact: row captured by feature_flags.js.

  • Early warning / validation: toolbar count assertion.

  • Mitigation: no data-feature-flag-* attributes on the row.

• Canonical components/API contracts touched: Feature Flags template/row layout, .switch, high_traffic_control.js; /api/prepare_high_traffic and /api/finalize_high_traffic unchanged.

Stage 2

Estimate: 0.07

• Goal: verify the release condition visually and record the outcome.

• Dependencies: Stage 1.

• Expected changes:

  • Small CSS fixes only if a theme or phone width misrenders.

  • docs/plans/ui_copy_audit_checklist.md: entry for the toggle cleanup.

  • This cycle's Step 4 summary and ledger closeout.

• Verification approach: real browser at 390px and 1280px, light/dark, Word97/Chicago/Midnight screenshots inspected; non-manager view shows no row; no horizontal overflow or console errors.

• Estimate details: driver: theme matrix; uncertainty: theme-specific switch styling.

• Risks or open questions:

  • Impact: misaligned or invisible switch in a theme.

  • Early warning / validation: screenshot per theme.

  • Mitigation: theme-scoped CSS adjustment reusing existing tokens.

• Canonical components/API contracts touched: theme styles, UI copy audit checklist.

Study checkpoint

Step 3 handoff, 2026-10-27 UTC: scope v3 (Option A), baseline 6e32cd093e76da88820388159de08c95e7bd9a0d, R1/P2, Claude Code single exposed rater. 0.20 projected total reaffirmed, independently judged (not the stage sum). Stage-work originals S1 0.10, S2 0.07; annotations, not summed or fractions. Additionally inspected the browser test and flag-row CSS. No implementation, measurement or accepted outcome. Awaiting Approved Step 3; planning and ledger remain uncommitted.

— Claude Code · 2026-10-27 UTC · Model: claude-opus-5-5