acceptance-prep

name	acceptance-prep
description	Run the full acceptance testing pre-flight pipeline for a feature. Use when asked to "run acceptance prep", "run the pre-flight pipeline", "prepare for acceptance testing", "run [ACCEPTANCE-PREP]", or when an [ACCEPTANCE-PREP] issue is open and ready to start. Orchestrates all four pipeline layers: demo agent, adversarial reviewer, real-data stress agent, synthesis agent. Produces the synthesis brief that the human uses to run the [ACCEPTANCE] session.
argument-hint	PR number or feature name (e.g. 42 or "report export feature")

Acceptance Pre-Flight Pipeline

Orchestrate the four-layer pre-flight pipeline for a feature and produce the synthesis brief the human needs to open the [ACCEPTANCE] session.

---

Prerequisites

Before starting:

• [IMPL] PR is merged
• Playwright suite passes against the current build
  • Run or confirm: make test / npx playwright test (or project equivalent)
  • If Playwright fails: stop here. Return the ticket to [IMPL].
• A real data snapshot is available (sanitized/anonymized)
  • If no snapshot exists: note this gap, proceed with Layer 1 and 2 only, and flag that Layer 3 was skipped and why

If the PR number is not provided and cannot be inferred, ask before proceeding.

---

Layer 1: Demo Agent

Apply the demo-agent persona.

Read the PR diff and CLAUDE.md. Produce ACCEPTANCE-DEMO.md covering:

• Every user-facing flow (entry point, happy path, expected output)
• Edge cases handled and not handled per flow
• Integration points and failure handling
• Assumptions made during implementation
• Three lowest-confidence areas

Post the demo document as a PR comment or artifact before proceeding to Layer 2.

---

Layer 2: Adversarial Reviewer

Apply the adversarial-reviewer persona.

Important: Treat this as a fresh context. Read ACCEPTANCE-DEMO.md and the PR diff. Do not carry forward the reasoning from Layer 1.

Find:

• Flows not demoed
• Assumptions that won't hold with real data
• Optimistic claims that can't be verified
• Edge cases in the "not handled" list that warrant reconsideration

Label all findings: [BLOCKING], [AMBIGUITY], [FALSE PRECISION], [SUGGESTION], [NIT].

After Layer 2:

• [BLOCKING] findings → open [BUG] tickets before proceeding to Layer 3
• [AMBIGUITY] findings → carry forward to Layer 4 (synthesis)
• Other findings → carry forward to Layer 4

---

Layer 3: Real Data Stress Agent

Apply the real-data-stress persona.

Run each flow from ACCEPTANCE-DEMO.md against the real data snapshot. Target specifically:

• Null or missing fields
• Long strings and special characters
• Boundary dates
• Orphaned references
• Unexpected record shapes

Produce the structured failure report with [CRITICAL], [DEGRADED], and [COSMETIC] classifications.

After Layer 3:

• [CRITICAL] failures → open [BUG] tickets before proceeding to Layer 4
• [DEGRADED] and [COSMETIC] findings → carry forward to Layer 4

---

Layer 4: Synthesis Agent

Apply the synthesis-agent persona.

Read all three prior outputs. Produce the human decisions brief:

• 2–4 items only
• Each item must require human judgment — not engineering judgment
• Scope questions, UX trade-offs, stakeholder calls, product decisions

Post the brief to the [ACCEPTANCE-PREP] issue as the final output.

---

If Blocking Issues Are Found and Fixed

When [BLOCKING] adversarial findings or [CRITICAL] real-data failures produce [BUG] tickets that are subsequently fixed, do not re-run the entire pipeline from Layer 0. Re-run only from the layer that found the blocking issue:

• [BLOCKING] from Layer 2 fixed → re-run Layer 2 (adversarial review) only, then re-run Layer 4 (synthesis) to update the brief
• [CRITICAL] from Layer 3 fixed → re-run Layer 3 (real data stress) only, then re-run Layer 4 (synthesis) to update the brief
• If the fix is large enough to plausibly affect earlier layers (e.g., a major flow was changed), re-run from Layer 1

Layer 0 (Playwright) does not need to re-run unless the bug fix touches existing tested flows — that is a judgment call, not a rule.

---

Output: [ACCEPTANCE-PREP] Completion Report

After all four layers, post the following to the [ACCEPTANCE-PREP] issue:

## [ACCEPTANCE-PREP] Complete — [Feature Name]

### Pipeline Status

- [x] Layer 0: Playwright — passed / [N] failures (return to [IMPL] if any)
- [x] Layer 1: Demo Agent — ACCEPTANCE-DEMO.md produced
- [x] Layer 2: Adversarial Review — [N] [BLOCKING], [N] [AMBIGUITY], [N] other
- [x] Layer 3: Real Data Stress — [N] [CRITICAL], [N] [DEGRADED], [N] [COSMETIC]
- [x] Layer 4: Synthesis — [N] human decisions identified

### Tickets Opened
[List any [BUG] tickets opened during pre-flight, with links]

### Human Decisions Brief
[Paste synthesis output here]

### Ready for [ACCEPTANCE]
- [ ] All [BLOCKING] adversarial findings have linked [BUG] tickets
- [ ] All [CRITICAL] real-data failures have linked [BUG] tickets
- [ ] Human decisions brief has been read

The [ACCEPTANCE-PREP] issue closes when this checklist is complete. The [ACCEPTANCE] issue can then open.