Task: Add T23 (bootstrap stub flow) and T24 (cost-ledger completeness) to the
testkit, bringing total from 27/27 to 29/29 PASS with 0 FAIL, 0 SKIP.
## What was changed
- `artifacts/api-server/src/routes/testkit.ts`:
- Updated audit-log comment block to document T23 + T24 additions.
- Inserted Test 23 after T22 (line ~654):
POST /api/bootstrap → assert 201 + bootstrapJobId present.
Guard on stubMode=true; SKIP if real DO mode (prevents hanging).
Stub-pay the paymentHash via /api/dev/stub/pay/:hash.
Poll GET /api/bootstrap/:id every 2s (20s timeout) until
state=provisioning or state=ready; assert message field present.
- Inserted Test 24 after T23:
Guarded on STATE_T6=complete (reuses completed job from T6).
GET /api/jobs/:id, extract costLedger.
Assert all 8 fields non-null: actualInputTokens, actualOutputTokens,
totalTokens, actualCostUsd, actualAmountSats, workAmountSats,
refundAmountSats, refundState.
Honest-accounting invariant: actualAmountSats <= workAmountSats.
refundAmountSats >= 0.
refundState must match ^(not_applicable|pending|paid)$.
## No deviations from task spec
- T23 guard logic matches spec exactly (stubMode check before poll).
- T24 fields match the 8 specified in task-24.md plus the invariants.
- No changes to bootstrap.ts or jobs.ts — existing routes already correct.
## Test run result
29/29 PASS, 0 FAIL, 0 SKIP (fresh server restart, rate-limit slots clean).
T23: state=provisioning in 1s. T24: actualAmountSats(179)<=workAmountSats(182),
refundAmountSats=3, refundState=pending.
031ca5a5c3