How I build, start to finish.

# <feature>: PRD lite

**Problem**: what's broken, for whom, and why now.
**Users / context**: who hits this, how often, in what situation.
**Goal**: the one outcome that means success.
**Out of scope**: what we are deliberately NOT doing.
**Approach**: the shape of the solution (a level up from implementation).
**Success metric**: how we'll know, stated before we build.
**Risks / unknowns**: what could make this wrong.
**Rollout**: how it ships, and how it rolls back.

# <decision>: ADR

**Status**: proposed | accepted | superseded
**Context**: the forces in play; what makes this a real choice.
**Options considered**: each with its honest tradeoffs.
**Decision**: what we chose.
**Rationale**: why this beat the alternatives.
**Consequences**: what this makes easy, and what it costs.

# <experiment>: note

**Hypothesis**: we believe X will cause Y, because Z.
**Method**: what we ran, on whom/what, how controlled.
**Measured**: the metric(s), defined up front.
**Result**: what actually happened (including "nothing").
**Decision**: ship / iterate / drop, and why.

# Release: <label>

- [ ] Tests + typecheck green
- [ ] `redact:check` clean (no secrets/hosts leaked)
- [ ] Customer value written (not just the diff)
- [ ] Known risks noted
- [ ] Rollback path exists and is understood
- [ ] Changelog / release note updated
- [ ] Verified in the real environment, not just locally

# <bug>

**Summary**: one sentence.
**Repro**: the shortest reliable steps.
**Expected vs actual**: what should happen; what does.
**Impact / severity**: who's affected, how badly.
**Environment**: where it shows up.

# <incident>: postmortem

**Impact**: what users/experience felt.
**Detection**: how we found out.
**Response**: what we did, in order.
**Root cause**: the real mechanism, not the symptom.
**Fix**: what changed, and how we verified it.
**Followup actions**: owner + status for each.
**Blameless note**: the systemic lesson, not the culprit.