AgentProof

Admin · Pilot run

End-to-end pilot run

The 15-step journey a customer/admin walks from opening Home through confirming the go/no-go verdict. Each step has a named proof source, an action route, an owner, and a last-checked timestamp. Steps with a verifier API path are system-verified — not button-only.

Open /admin/pilot-pack →Open /admin/pilot-launch (R38) →Open /admin/pilot-readiness →Open /system-health →

Pilot verification result (R35)

No verification result on record

No pilot verification result on record. Run the one-click verification on /system-health to confirm persistence works end-to-end.

Why pilot verification matters

Pilot verification is the product's own proof that Supabase persistence works end-to-end for the signed-in customer. Every readiness surface consumes the verdict — pilot readiness flips Ready only after this passes.

What it verifies

  • Supabase session exists for the signed-in customer.
  • Supabase is configured (URL + anon key) on this deploy.
  • manual_environments table accepts upsert + read-back.
  • manual_agents table accepts upsert + read-back.
  • review_snapshots table accepts upsert + read-back.
  • report_history table accepts upsert + read-back.

Next action: Open /system-health and click Run pilot verification.

Run pilot verification →

The cached run is a localStorage cache of the/api/system-health/pilot-verification-run endpoint response — never proof on its own. The endpoint return value is the authoritative source.

R37 / S34BI · Auto-executable pilot run

Pilot run executor

Executes the 11 system-verifiable pilot steps. Per-step proof source, verifier API, blocker, and exact next action.

Verdict · not ready

Pilot run has not been executed yet. Click 'Run pilot check now'.

Next action: Run pilot check nowOpen /admin/pilot-run

Run id
—
Started
—
Completed
—
Passed
0
Failed
0
Blocked
0

    R37 / S34BI · Live configuration completion

    Configuration completion

    2 pilot-blocking configuration item(s) outstanding.

    Variable names only · never echoes secret values.

    Configured
    3
    Inferred
    0
    Mismatch
    0
    Missing
    3
    Optional
    2

    1. Public base URL

      missing

      Not configured and cannot be inferred from request headers.

      Variable name
      AGENTPROOF_PUBLIC_BASE_URL
      Consumed by
      GitHub Actions scheduled radar workflow; Email link rendering; R34 Radar schedule validator

      Next action: Set AGENTPROOF_PUBLIC_BASE_URL on RailwayOpen /admin/setup

    2. Supabase public URL

      missing

      Supabase URL is not configured.

      Variable name
      NEXT_PUBLIC_SUPABASE_URL
      Consumed by
      Auth + persistence; R12 system health probe

      Next action: Set NEXT_PUBLIC_SUPABASE_URL on RailwayOpen /admin/setup

    3. Supabase publishable (anon) key

      missing

      Supabase anon key is not configured.

      Variable name
      NEXT_PUBLIC_SUPABASE_ANON_KEY
      Consumed by
      Auth + persistence; R12 system health probe

      Next action: Set NEXT_PUBLIC_SUPABASE_ANON_KEY on RailwayOpen /admin/setup

    4. Admin tools

      optional missing

      Admin tools disabled — pilot admin must enable on Railway to access /admin/* routes.

      Variable name
      AGENTPROOF_ADMIN_TOOLS_ENABLED
      Consumed by
      /admin/* routes

      Next action: Set AGENTPROOF_ADMIN_TOOLS_ENABLED=true on RailwayOpen /admin/setup

    5. Radar preview (public)

      optional missing

      Radar preview is hidden behind the admin flag.

      Variable name
      AGENTPROOF_RADAR_PREVIEW_PUBLIC_ENABLED
      Consumed by
      /learn/ai-landscape-radar

      Next action: Optional: set AGENTPROOF_RADAR_PREVIEW_PUBLIC_ENABLED=true to publish the previewOpen /learn/ai-landscape-radar

    6. Scheduled radar workflow

      configured

      GitHub Actions workflow file present in repo.

      Variable name
      .github/workflows/radar-scheduled-run.yml
      Consumed by
      R33 scheduled radar runner

      Next action: No action neededOpen /admin/intelligence-ops

    7. GitHub Actions repository secret

      configured

      Repo secret AGENTPROOF_PUBLIC_BASE_URL is documented in workflow.

      Variable name
      AGENTPROOF_PUBLIC_BASE_URL (GitHub Actions secret)
      Consumed by
      .github/workflows/radar-scheduled-run.yml

      Next action: Verify the value on GitHub → Settings → Secrets

    8. Documentation: where secrets are consumed

      configured

      AGENTPROOF_PUBLIC_BASE_URL is consumed by the workflow at .github/workflows/radar-scheduled-run.yml and by scripts/run_radar_scheduled_check.cjs.

      Variable name
      (see workflow + scripts)
      Consumed by
      .github/workflows/radar-scheduled-run.yml; scripts/run_radar_scheduled_check.cjs

      Next action: No action neededOpen /admin/intelligence-ops

    Pilot journey state (R36)

    2 dimension(s) blocked. Address the named blocker before continuing.

    0 system-verified·0 user-attested·0 sample/demo·2 blocked·6 not started

    • Setup

      Blocked

      Setup wizard reports one or more critical items missing or failing.

      Evidence source: /api/admin/setup-status returns overall_state === 'all_critical_configured'.

      Store: Computed at render

      Blocker: Open /admin/setup and address every critical item.

      Open setup wizard →

    • Verification

      Not started

      R31 pilot verification has not been run for this session.

      Evidence source: R31 cached run from /api/system-health/pilot-verification-run.

      Store: Local cache (UI tracking only)

      Blocker: Open /system-health and click Run pilot verification.

      Run pilot verification →

    • Workspace

      Not started

      Customer is not signed in. Workspace is empty.

      Evidence source: SSR session detection + manual_agents row presence via the verify endpoint.

      Store: Computed at render

      Blocker: Pilot customer signs in.

      Open workspace →

    • Manual agent

      Not started

      No manual agent verified yet for the signed-in user.

      Evidence source: /api/workspace/manual-agent/verify confirms a real row.

      Store: Computed at render

      Blocker: Open /workspace/manual-agent/new and submit the form.

      Add manual agent →

    • Review

      Not started

      No committed review snapshot for any agent yet.

      Evidence source: review_snapshots Supabase upsert + read-back via R31 verification.

      Store: Computed at render

      Blocker: Open /workspace/readiness and walk the review.

      Open the readiness review →

    • Report

      Not started

      No report rows visible yet for the signed-in user.

      Evidence source: /api/reports/verification returns supabase_real_rows.

      Store: Computed at render

      Blocker: Complete a review to generate the first report.

      Open the reports centre →

    • AI Radar

      Blocked

      Radar schedule artefact ships but endpoint is not reachable (likely missing AGENTPROOF_PUBLIC_BASE_URL).

      Evidence source: /api/radar/schedule-validator returns configured_runs_clean.

      Store: Computed at render

      Blocker: Add AGENTPROOF_PUBLIC_BASE_URL secret on the deploy.

      Open intelligence ops →

    • Pilot readiness

      Not started

      Pilot readiness has not flipped to Ready yet.

      Evidence source: R34 PilotReadinessFinalGating verdict === 'ready_for_controlled_private_pilot'.

      Store: Computed at render

      Blocker: Resolve remaining blockers on /admin/pilot-readiness.

      Open pilot readiness →

    The same journey state shape renders on /workspace, /system-health, /admin/pilot-pack, /admin/pilot-readiness, and /admin/pilot-run — one consistent view across surfaces.

    Pilot run journey (R36)

    8 step(s) blocked or failed. Address the named blocker before continuing.

    0 system-verified·3 user-attested·4 pending·8 blocked/failed

    1. Step 1. Open Home

      User attestedOwner: Pilot customer

      Pilot customer opens the public home page on the deploy URL.

      Proof required: Customer opens / and confirms the landing renders.

      Proof source: Visual confirmation — no API verifier.

      Last checked: never

      Open / →

    2. Step 2. Sign in

      Not startedOwner: Pilot customer

      Pilot customer signs in with the invited identity.

      Proof required: SSR shell reports a signed-in Supabase session for the customer.

      Proof source: Supabase auth.getSession() inside the server shell or R31 endpoint.

      Last checked: never

      Blocker: Customer signs in.

      Open /auth/login →

    3. Step 3. Run pilot verification

      BlockedOwner: Pilot customer

      Click Run pilot verification on /system-health. The R31 endpoint writes + reads back 4 records.

      Proof required: R31 cached run verdict === 'passed' with all step outcomes verified.

      Proof source: /api/system-health/pilot-verification-run. · Verifier: /api/system-health/pilot-verification-run

      Last checked: never

      Blocker: Customer signs in first.

      Open /system-health#one-click-pilot-verification →

    4. Step 4. Open the workspace

      BlockedOwner: Pilot customer

      Customer opens /workspace and confirms the readiness journey + pilot mode panels render.

      Proof required: Workspace renders without dead ends and shows the verification result consumer.

      Proof source: Customer-attested. Workspace pilot mode panel reflects the verification state.

      Last checked: never

      Blocker: Customer signs in first.

      Open /workspace →

    5. Step 5. Add a manual non-Microsoft agent

      BlockedOwner: Pilot customer

      Open /workspace/manual-agent/new, submit the form, and confirm the verified success panel.

      Proof required: At least one verified manual_agents row exists for the signed-in user.

      Proof source: /api/workspace/manual-agent/verify. · Verifier: /api/workspace/manual-agent/verify

      Last checked: never

      Blocker: Submit the manual non-Microsoft agent form.

      Open /workspace/manual-agent/new →

    6. Step 6. Start the readiness review

      BlockedOwner: Pilot customer

      Open /workspace/readiness for the new agent and walk the first answers.

      Proof required: Review snapshot present (in-progress or committed) for the agent.

      Proof source: review_snapshots Supabase row.

      Last checked: never

      Blocker: Add a manual agent first.

      Open /workspace/readiness →

    7. Step 7. Save answers

      BlockedOwner: Pilot customer

      Commit at least one review snapshot — placeholders and 'Legacy answer unavailable' do NOT count.

      Proof required: Review snapshot reports committed === true with real answers (not placeholders).

      Proof source: review_snapshots upsert + read-back via R31. · Verifier: /api/system-health/pilot-verification-run

      Last checked: never

      Blocker: Walk the review and save real answers.

      Open /workspace/readiness →

    8. Step 8. Generate the report

      BlockedOwner: AgentProof engine

      After committing the review, the report row lands in report_history.

      Proof required: At least one report_history row visible to the signed-in user.

      Proof source: /api/reports/verification. · Verifier: /api/reports/verification

      Last checked: never

      Blocker: Commit a review to trigger report generation.

      Open /reports →

    9. Step 9. Open the reports centre

      BlockedOwner: Pilot customer

      /reports lists at least one real report with the four version stamps.

      Proof required: ReportsCentreDataPanel renders real rows.

      Proof source: Customer-attested via /reports + /api/reports/verification. · Verifier: /api/reports/verification

      Last checked: never

      Blocker: Generate the first report first.

      Open /reports →

    10. Step 10. Export / print the report

      In progressOwner: Pilot customer

      Use the export / print action on the report row or detail page.

      Proof required: Export action visible on at least one report and the print layout is clean.

      Proof source: Customer-attested via the export/print action on the report card.

      Last checked: never

      Open /reports#export →

    11. Step 11. Open AI Radar

      User attestedOwner: Pilot customer

      /learn/ai-landscape-radar surfaces the customer-safe operating loop.

      Proof required: Radar Learn page renders without a fake real-time monitoring claim.

      Proof source: Customer-attested.

      Last checked: never

      Open /learn/ai-landscape-radar →

    12. Step 12. Run or check AI Radar

      In progressOwner: Product admin

      Open /admin/intelligence-ops and confirm the schedule validator + run history.

      Proof required: At least one radar run (scheduled or manual) is recorded.

      Proof source: /api/radar/schedule-validator + run history. · Verifier: /api/radar/schedule-validator

      Last checked: never

      Blocker: Configure AGENTPROOF_PUBLIC_BASE_URL or click Run on /admin/intelligence-ops.

      Open /admin/intelligence-ops →

    13. Step 13. Review AI Radar impact

      BlockedOwner: Product admin

      Confirm whether approved Radar updates affect the generated report.

      Proof required: Adaptation governance panel + reports newer-intelligence warning aligned.

      Proof source: R33 governance + R34 reports verification view.

      Last checked: never

      Blocker: Record a radar run first.

      Open /admin/intelligence-ops →

    14. Step 14. Open the pilot readiness verdict

      User attestedOwner: Product admin

      /admin/pilot-readiness consumes setup + verification + reports + radar + trust.

      Proof required: PilotReadinessFinalGatingPanel renders with consumed validators visible.

      Proof source: Customer-attested via /admin/pilot-readiness.

      Last checked: never

      Open /admin/pilot-readiness →

    15. Step 15. Confirm go / no-go verdict

      In progressOwner: Product admin

      Record the go/no-go decision against the stop/go criteria.

      Proof required: Final-gating verdict === 'ready_for_controlled_private_pilot'.

      Proof source: R34 PilotReadinessFinalGating view consumed validators.

      Last checked: never

      Blocker: Resolve remaining blockers on /admin/pilot-readiness.

      Open /admin/pilot-readiness →

    Steps backed by a verifier API path are derived from the named endpoint, not from a button click. The runner refuses to mark such a step system-verified unless its verifier has confirmed it.

    Route smoke inventory (R36)

    All 17 routes in the smoke inventory resolve to a real page.

    17 ok·0 missing·0 audit blockers
    • /OK

      Route / resolves to app/page.tsx.

      Source: app/page.tsx · Shell: public

    • /learnOK

      Route /learn resolves to app/learn/page.tsx.

      Source: app/learn/page.tsx · Shell: public

    • /demoOK

      Route /demo resolves to app/demo/page.tsx.

      Source: app/demo/page.tsx · Shell: public

    • /trialOK

      Route /trial resolves to app/trial/page.tsx.

      Source: app/trial/page.tsx · Shell: public

    • /workspaceOK

      Route /workspace resolves to app/workspace/page.tsx.

      Source: app/workspace/page.tsx · Shell: workspace

    • /workspace/manual-agent/newOK

      Route /workspace/manual-agent/new resolves to app/workspace/manual-agent/new/page.tsx.

      Source: app/workspace/manual-agent/new/page.tsx · Shell: workspace

    • /reportsOK

      Route /reports resolves to app/reports/page.tsx.

      Source: app/reports/page.tsx · Shell: public

    • /learn/ai-landscape-radarOK

      Route /learn/ai-landscape-radar resolves to app/learn/ai-landscape-radar/page.tsx.

      Source: app/learn/ai-landscape-radar/page.tsx · Shell: public

    • /system-healthOK

      Route /system-health resolves to app/system-health/page.tsx.

      Source: app/system-health/page.tsx · Shell: public

    • /admin/setupOK

      Route /admin/setup resolves to app/admin/setup/page.tsx.

      Source: app/admin/setup/page.tsx · Shell: admin

    • /admin/pilot-readinessOK

      Route /admin/pilot-readiness resolves to app/admin/pilot-readiness/page.tsx.

      Source: app/admin/pilot-readiness/page.tsx · Shell: admin

    • /admin/pilot-packOK

      Route /admin/pilot-pack resolves to app/admin/pilot-pack/page.tsx.

      Source: app/admin/pilot-pack/page.tsx · Shell: admin

    • /admin/pilot-runOK

      Route /admin/pilot-run resolves to app/admin/pilot-run/page.tsx.

      Source: app/admin/pilot-run/page.tsx · Shell: admin

    • /admin/intelligence-opsOK

      Route /admin/intelligence-ops resolves to app/admin/intelligence-ops/page.tsx.

      Source: app/admin/intelligence-ops/page.tsx · Shell: admin

    • /admin/pilot-launchOK

      Route /admin/pilot-launch resolves to app/admin/pilot-launch/page.tsx.

      Source: app/admin/pilot-launch/page.tsx · Shell: admin

    • /admin/customer-acceptanceOK

      Route /admin/customer-acceptance resolves to app/admin/customer-acceptance/page.tsx.

      Source: app/admin/customer-acceptance/page.tsx · Shell: admin

    • /pilot/sessionOK

      Route /pilot/session resolves to app/pilot/session/page.tsx.

      Source: app/pilot/session/page.tsx · Shell: public

    Pilot session — live (run observer)

    storage: local_cache_only

    Sign in for Supabase-backed persistence.

    AI Radar — operating state visible to run

    Approved-source watchlist only. No broad crawling. No company probing. Live mode requires AGENTPROOF_RADAR_LIVE_FETCH=true server-side; otherwise the run is dry-run and clearly labelled.

    No radar run cached in this browser yet. Run AI Radar check now to record one.

    Storage: local_cache_only — this cache is per browser. Supabase-backed persistence is a separate piece.