# Hatchways Agentic Mode Rollout Checklist

Convert the live proof packet into a concrete limited-pilot checklist for a Hatchways product or partnerships owner, without implying a current integration or partnership.

## Recommended Scope

One internal Hatchways-style GitHub assessment, one reviewer, one candidate-style dry run, and no production customer traffic until the evidence packet passes the acceptance bar.

## Owners

- Hatchways product owner: Choose the assessment repo/task, reviewer, and pilot acceptance bar.
- Hatchways engineering owner: Add the GitHub Actions fallback or provide GitHub App install access when configured.
- HotTea owner: Provision the candidate workspace, verify evidence ingestion, and package the reviewer packet.
- Reviewer: Review the packet beside Hatchways hidden grading and answer the pilot questions.

## Fourteen-Day Plan

### Day 0: Pilot scoping

- Action: Pick one repo-based assessment and confirm no real candidate/customer traffic is included.
- Proof: https://hottea.ai/hatchways/decision.md

### Day 1: Adapter setup

- Action: Attach the GitHub Actions fallback to the pilot repo or configure the real GitHub App only if GITHUB_APP_INSTALL_URL is available.
- Proof: https://hottea.ai/hatchways/integration.md

### Day 2: Candidate dry run

- Action: Run one candidate-style session through the provisioned VM, task packet, ae-claude/ae-codex wrappers, tests, ae-finalize, and final submission.
- Proof: https://hottea.ai/hatchways/pilot.md

### Day 3: Reviewer calibration

- Action: Have one reviewer inspect the packet using correctness, process, AI collaboration, anomaly, and handoff dimensions.
- Proof: https://hottea.ai/hatchways/evaluation.md

### Days 4-7: Internal comparison

- Action: Compare the HotTea packet to the normal Hatchways reviewer surface for the same task.
- Proof: https://hottea.ai/hatchways/business-case.md

### Days 8-14: Go/no-go

- Action: Decide whether the packet improves reviewer confidence enough to justify GitHub App configuration and a customer-safe pilot.
- Proof: https://hottea.ai/hatchways/readiness.md

## Acceptance Bar

- Reviewer can answer the five pilot questions without a live candidate debrief.
- Packet shows hidden-test status, public test output, git diff/snapshot evidence, Claude/Codex session exports, and VM audit events.
- Reviewer can separate procedural AI prompts from substantive candidate reasoning.
- Current blockers and non-claims are clear before any customer or procurement conversation.

## Rollback Plan

- Remove the GitHub Actions fallback from the pilot repo.
- Expire or teardown the candidate VM.
- Keep the generated reviewer packet only as internal pilot evidence.
- Do not send customer-facing claims if the acceptance bar fails.

## Blocked Until Account Work

- Real GitHub App creation and install URL configuration.
- Official Hatchways partnership approval.
- Enterprise procurement artifacts such as SOC 2, DPA, SSO, retention SLA, and vendor security review.

## Not Claimed

- No current Hatchways partnership.
- No production customer rollout.
- No real GitHub App before GITHUB_APP_INSTALL_URL is configured.
- No perfect anti-cheat or outside-AI prevention.
- No automated hiring decision.

## Proof URLs

- Rollout JSON: https://hottea.ai/hatchways/rollout.json
- Decision brief: https://hottea.ai/hatchways/decision.md
- Integration guide: https://hottea.ai/hatchways/integration.md
- Pilot kit: https://hottea.ai/hatchways/pilot.md
- Evaluation rubric: https://hottea.ai/hatchways/evaluation.md
- Readiness scorecard: https://hottea.ai/hatchways/readiness.md
- Sample reviewer packet: https://hottea.ai/sample-report