[wizard-ci-bot]

Automated wizard CI run

Source: scheduled
Trigger ID: cc5471b
App: react-vite/court-booking
App directory: apps/react-vite/court-booking
Workbench branch: wizard-ci-cc5471b-react-vite-court-booking
Wizard branch: main
Context Mill branch: main
PostHog (MCP) branch: master
Timestamp: 2026-10-28T17:12:07.238Z
Duration: 162.4s

YARA Scanner

✓ 68 tool calls scanned, 0 violations detected

No violations: ✓ 68 clean scans

[wizard-ci-bot]

Now I have all the context needed. Let me evaluate the PR.

---

PR Evaluation Report

Summary

The replay-vision wizard installed posthog-js and initialized it with env-read credentials in the Vite React court-booking app (previously uninstrumented). It wrote a run report listing three planned scanners. However, the scanners were not actually created — the report states credential scopes were insufficient — and consequently no credit estimates, sampling rates, or model choices are verifiable. The scanner descriptions are well-tailored to the app's vocabulary and routes.

Files changed	Lines added	Lines removed
5	+55	-0

Confidence score: 5/5 🧙

• Scanners not created: The report documents intended scanners but none were actually created due to claimed missing MCP credential scopes. The rubric expects three scanners to exist, not three intentions. [CRITICAL]
• No credit estimates: Monthly credit spend is listed as "Not estimated" for all three scanners, failing the cost discipline requirement. [CRITICAL]
• Sampling rates and model not verifiable: The report mentions "intended" sampling of 0.5 and 0.1 and gemini-3-flash-preview, but since no scanners were created, these values cannot be verified against the rubric's requirements. [MEDIUM]

---

File changes

Filename	Score	Description
.env.example	4/5	Adds VITE_POSTHOG_KEY and VITE_POSTHOG_HOST env vars — correct for Vite
.gitignore	5/5	Ignores .env to prevent secret leaks
package.json	5/5	Adds posthog-js dependency
posthog-replay-vision-report.md	2/5	Report exists but scanners are deferred, no credit estimates
src/main.tsx	5/5	Clean, minimal posthog-js init with env vars and defaults: '2026-08-30'

---

App sanity check ⚠️

Criteria	Result	Description
App builds and runs	Yes	posthog-js added as dependency, init is guarded by env vars so app works without them
Preserves existing env vars & configs	Yes	No existing configs modified, only additions
No syntax or type errors	Yes	TypeScript is clean, proper Vite import.meta.env usage
Correct imports/exports	Yes	posthog-js imported correctly as default import
Minimal, focused changes	Yes	Only PostHog SDK install and init — no identify, capture, dashboards, or other products
Pre-existing issues	None	—

Other completed criteria

• .env.example documents the two required environment variables
• .gitignore prevents .env from being committed
• Init uses defaults: '2026-08-30' per current docs
• Error guards in dev mode give actionable messages when env vars are missing
• No app routes, components, or behavior changed

---

PostHog implementation ✅

Criteria	Result	Description
PostHog SDKs installed	Yes	posthog-js ^1.441.2 added to dependencies
PostHog client initialized	Yes	posthog.init() with env-read key and host, defaults: '2026-08-30'
capture()	N/A	Replay-vision wizard is not expected to add capture calls
identify()	N/A	Replay-vision wizard is not expected to add identify calls
Error tracking	N/A	Not in scope for replay-vision
Reverse proxy	N/A	Not in scope for replay-vision

Other completed criteria

• Init is minimal — no unnecessary options, no identify, no capture, no dashboards
• Follows the existing code style (TypeScript, Vite patterns)
• Environment variable names use VITE_ prefix per Vite convention

---

PostHog insights and events ⚠️

Filename	PostHog events	Description
src/main.tsx	Autocapture (default)	posthog-js initialized with defaults; autocapture, pageviews, and session replay are enabled by default

Issues

• Scanners not created: The report lists three intended scanners but explicitly states none were created due to insufficient MCP credential scopes. The wizard should have created these scanners as part of its run. [CRITICAL]
• No credit estimates provided: All three scanners show "Not estimated" for monthly credits, violating the cost discipline requirement. [CRITICAL]
• Sampling and model unverifiable: Intended sampling rates (0.5 breakage, 0.1 summarizer) and model (gemini-3-flash-preview) are mentioned in prose but cannot be verified since no scanners exist. The frustration monitor's intended sampling is not stated. [MEDIUM]

Other completed criteria

• Scanner names are product-specific: "Booking failures", "Padel booking frustration", "Padel booking session summaries" — not generic
• Breakage monitor scopes to /book/ and /confirmation — matches real app routes
• Frustration monitor is gated on `` with no additional URL filters
• Summarizer is unscoped (intended to sample across the product)
• Prompts use booking vocabulary: padel courts, time slots, bookings, confirming a booking
• No file paths, secrets, or instruction-shaped text in scanner descriptions

---

Reviewed by wizard workbench PR evaluator