[pyjeebz]

Stacked on #9. The base is #9's branch, so the diff shows only this slice. It gets retargeted to main once #9 merges.

Journey results now show up wherever results show up:

• HTML report (--html, show --html): a Journeys section with each journey's verdict (survives, band, the step that breaks first, why), the combined run with any interference, and the --expect answer.
• prescale compare (terminal, --json, --markdown) diffs each journey by name. In a journey run, a drop past the threshold in any journey is a regression, even when it isn't the weakest journey. That matches the settled M4 rule: confidence labels a regression, it doesn't gate it. A journey that held everything it was tested to, below the old level, is inconclusive, not a drop. New, removed and stale journeys are listed, never counted as regressions.
• MCP run summaries list each journey compactly (status, survives, the step that breaks, why), plus the combined run's interference and the --expect answer.

How it was verified

Two real runs of browse + checkout: one against the example shop, one against a copy with a 10-connection pool instead of 25 (like a bad deploy).

$ prescale compare --markdown
| journey | old | new | change | breaks at |
| browse | 200+ | 200+ | +0% | — |
| checkout | ~20 | ~10 | -50% 🔴 | place order |
$ prescale compare --fail-on-regression    → exit 1

• The HTML report from a real two-journey run with interference shows the journeys table, the "Together" box and "Expected ~60 users: no". I checked a screenshot at 1000px.
• The MCP summary of that run: culprit_journey: checkout, breaks_at_step: "3. place order (POST /api/carts/{{cart_id}}/order)", together.interference: {users_at_break: 15, held_alone: 20}, expect.meets: false. About 1.6 KB.
• Unit tests cover:
  • a drop in a journey that isn't the weakest still regresses
  • a confident overall gain doesn't make a noisy journey drop confident
  • inconclusive and held rows
  • | in names escaped in markdown
  • HTML escaping of every user-controlled string
  • URL runs unchanged (no journeys section or field)
• ruff check cli/src: clean · pytest cli/tests -q: 314 passed.

Risk

Tier: risky · Public surface: compare --json gains journeys rows; --fail-on-regression fails on a drop past the threshold in any journey of a journey run; the MCP summary gains journeys, together and expect; the HTML report gains a section. Rollback: revert. These are renderers and comparison logic only; nothing stored or uploaded changes.

Fresh-context review: ok to ship. All its should-fixes are taken:

• confidence now comes only from drops
• the docs said "confident drops" gate, but the code (correctly, per M4) gates any drop past the threshold, so the docs were fixed
• inconclusive and held journeys are labelled as such
• | is escaped in markdown cells
• it injected HTML into every journey field: all escaped

Notes for review

prescale ci and the GitHub Action still run URL mode with their own comparison. Journeys in CI (with --stale warn|fail and a per-journey PR comment) is the next PR, J1.4b.

---

---

Generated by Claude Code

[Copilot]

Copilot was unable to review this pull request because the user who requested the review has insufficient remaining quota. Try requesting a review at a lower effort level.