=== scenario: empty
# Status — demo

**GENERATED** by `bash harness/engine/core/orchestrator/orchestrate.sh status` (STORY-010 T-CDL-006).
Do not hand-edit — rerun the command to refresh. Hand-written history lives in git.
**As of (newest metrics event):** (no metrics recorded)

## Queue

| | Count |
|---|---|
| Done |  |
| Pending |  |
| Failed (re-queueable) |  |
| Skipped (deps/filter) |  |

**Next task (priority order):** (queue empty)

## Lanes in flight

Elapsed is measured against the **newest metrics event**, not the clock — the
same instant the "as of" line reports. That is what keeps two `status` runs on
unchanged state byte-identical (STORY-010 AC-3). Spend is the runner-reported
figure for that task, from `runner_cost` events; see the Spend caveat below.

| Lane | Task | Spec | Elapsed | Spend | State |
|---|---|---|---|---|---|
| — | (no lane in flight) | — | — | — | — |

**Lane pool** (cap 3, max 1/story):

  - every configured lane is available

## Tuning numbers

Derived from `metrics.jsonl` alone (`orchestrate.sh lane-metrics`). They
inform a human choosing `max_concurrent_lanes`; the scheduler never tunes
itself.

| Metric | Value |
|---|---|
| Lane utilisation | — |
| Semaphore wait | — |
| Merge-conflict rate | — |

## Stories

| Story | Status | Tasks done/total |
|---|---|---|
| (no stories) | — | — |

## Bugs

- Open: 0
- Escalated lanes: none

## QA health

| Metric | Value |
|---|---|
| QA verdicts recorded | 0 |
| QA pass rate | — |
| Bug-escape rate | — |
| Mean task wall-time | — |

## Spend

**Runner-reported, advisory, and a FLOOR.** Every figure below is what the agent
client itself said an invocation cost. Plane B has no gateway meter in front of
the client, so nothing corroborates these numbers and nothing bills from them —
they are provenance, not billing truth. They are SPEND, not a cap: neither the
provider account's limit nor any `runners.<role>.budget_usd` this project has
configured. Dispatches on an adapter that cannot report cost are excluded
entirely rather than counted as $0.00, so the true spend is **at least** this.

| Metric | Value |
|---|---|
| Total (all recorded runs) | $0.0000 |
| Priced dispatches | 0 |
| Dispatches that reported nothing | 0 |
| Degradations (adapter cannot price a role) | 0 |

| Story | Spend | Priced dispatches |
|---|---|---|
| (no priced dispatches) | — | — |

## Reference

- Queue: [harness/state/queue.txt](queue.txt)
- Metrics: [harness/state/metrics.jsonl](metrics.jsonl)
- Blockers: [harness/state/blockers.md](blockers.md)
rc=0
--- current-status.md:
# Status — demo

**GENERATED** by `bash harness/engine/core/orchestrator/orchestrate.sh status` (STORY-010 T-CDL-006).
Do not hand-edit — rerun the command to refresh. Hand-written history lives in git.
**As of (newest metrics event):** (no metrics recorded)

## Queue

| | Count |
|---|---|
| Done |  |
| Pending |  |
| Failed (re-queueable) |  |
| Skipped (deps/filter) |  |

**Next task (priority order):** (queue empty)

## Lanes in flight

Elapsed is measured against the **newest metrics event**, not the clock — the
same instant the "as of" line reports. That is what keeps two `status` runs on
unchanged state byte-identical (STORY-010 AC-3). Spend is the runner-reported
figure for that task, from `runner_cost` events; see the Spend caveat below.

| Lane | Task | Spec | Elapsed | Spend | State |
|---|---|---|---|---|---|
| — | (no lane in flight) | — | — | — | — |

**Lane pool** (cap 3, max 1/story):

  - every configured lane is available

## Tuning numbers

Derived from `metrics.jsonl` alone (`orchestrate.sh lane-metrics`). They
inform a human choosing `max_concurrent_lanes`; the scheduler never tunes
itself.

| Metric | Value |
|---|---|
| Lane utilisation | — |
| Semaphore wait | — |
| Merge-conflict rate | — |

## Stories

| Story | Status | Tasks done/total |
|---|---|---|
| (no stories) | — | — |

## Bugs

- Open: 0
- Escalated lanes: none

## QA health

| Metric | Value |
|---|---|
| QA verdicts recorded | 0 |
| QA pass rate | — |
| Bug-escape rate | — |
| Mean task wall-time | — |

## Spend

**Runner-reported, advisory, and a FLOOR.** Every figure below is what the agent
client itself said an invocation cost. Plane B has no gateway meter in front of
the client, so nothing corroborates these numbers and nothing bills from them —
they are provenance, not billing truth. They are SPEND, not a cap: neither the
provider account's limit nor any `runners.<role>.budget_usd` this project has
configured. Dispatches on an adapter that cannot report cost are excluded
entirely rather than counted as $0.00, so the true spend is **at least** this.

| Metric | Value |
|---|---|
| Total (all recorded runs) | $0.0000 |
| Priced dispatches | 0 |
| Dispatches that reported nothing | 0 |
| Degradations (adapter cannot price a role) | 0 |

| Story | Spend | Priced dispatches |
|---|---|---|
| (no priced dispatches) | — | — |

## Reference

- Queue: [harness/state/queue.txt](queue.txt)
- Metrics: [harness/state/metrics.jsonl](metrics.jsonl)
- Blockers: [harness/state/blockers.md](blockers.md)
=== scenario: no-ledger
# Status — demo

**GENERATED** by `bash harness/engine/core/orchestrator/orchestrate.sh status` (STORY-010 T-CDL-006).
Do not hand-edit — rerun the command to refresh. Hand-written history lives in git.
**As of (newest metrics event):** (no metrics recorded)

## Queue

| | Count |
|---|---|
| Done | 1 |
| Pending | 1 |
| Failed (re-queueable) | 1 |
| Skipped (deps/filter) | 2 |

**Next task (priority order):** d

## Lanes in flight

Elapsed is measured against the **newest metrics event**, not the clock — the
same instant the "as of" line reports. That is what keeps two `status` runs on
unchanged state byte-identical (STORY-010 AC-3). Spend is the runner-reported
figure for that task, from `runner_cost` events; see the Spend caveat below.

| Lane | Task | Spec | Elapsed | Spend | State |
|---|---|---|---|---|---|
| 1 | T-K1 | docs/k1.md#t-k1 | — | no priced dispatch yet | running |

**Lane pool** (cap 3, max 1/story):

  - lane 1: DISABLED — reason 1

## Tuning numbers

Derived from `metrics.jsonl` alone (`orchestrate.sh lane-metrics`). They
inform a human choosing `max_concurrent_lanes`; the scheduler never tunes
itself.

| Metric | Value |
|---|---|
| Lane utilisation | — |
| Semaphore wait | — |
| Merge-conflict rate | — |

## Stories

| Story | Status | Tasks done/total |
|---|---|---|
| (no stories) | — | — |

## Bugs

- Open: 0
- Escalated lanes: none

## QA health

| Metric | Value |
|---|---|
| QA verdicts recorded | 0 |
| QA pass rate | — |
| Bug-escape rate | — |
| Mean task wall-time | — |

## Spend

**Runner-reported, advisory, and a FLOOR.** Every figure below is what the agent
client itself said an invocation cost. Plane B has no gateway meter in front of
the client, so nothing corroborates these numbers and nothing bills from them —
they are provenance, not billing truth. They are SPEND, not a cap: neither the
provider account's limit nor any `runners.<role>.budget_usd` this project has
configured. Dispatches on an adapter that cannot report cost are excluded
entirely rather than counted as $0.00, so the true spend is **at least** this.

| Metric | Value |
|---|---|
| Total (all recorded runs) | $0.0000 |
| Priced dispatches | 0 |
| Dispatches that reported nothing | 0 |
| Degradations (adapter cannot price a role) | 0 |

| Story | Spend | Priced dispatches |
|---|---|---|
| (no priced dispatches) | — | — |

## Reference

- Queue: [harness/state/queue.txt](queue.txt)
- Metrics: [harness/state/metrics.jsonl](metrics.jsonl)
- Blockers: [harness/state/blockers.md](blockers.md)
rc=0
--- current-status.md:
# Status — demo

**GENERATED** by `bash harness/engine/core/orchestrator/orchestrate.sh status` (STORY-010 T-CDL-006).
Do not hand-edit — rerun the command to refresh. Hand-written history lives in git.
**As of (newest metrics event):** (no metrics recorded)

## Queue

| | Count |
|---|---|
| Done | 1 |
| Pending | 1 |
| Failed (re-queueable) | 1 |
| Skipped (deps/filter) | 2 |

**Next task (priority order):** d

## Lanes in flight

Elapsed is measured against the **newest metrics event**, not the clock — the
same instant the "as of" line reports. That is what keeps two `status` runs on
unchanged state byte-identical (STORY-010 AC-3). Spend is the runner-reported
figure for that task, from `runner_cost` events; see the Spend caveat below.

| Lane | Task | Spec | Elapsed | Spend | State |
|---|---|---|---|---|---|
| 1 | T-K1 | docs/k1.md#t-k1 | — | no priced dispatch yet | running |

**Lane pool** (cap 3, max 1/story):

  - lane 1: DISABLED — reason 1

## Tuning numbers

Derived from `metrics.jsonl` alone (`orchestrate.sh lane-metrics`). They
inform a human choosing `max_concurrent_lanes`; the scheduler never tunes
itself.

| Metric | Value |
|---|---|
| Lane utilisation | — |
| Semaphore wait | — |
| Merge-conflict rate | — |

## Stories

| Story | Status | Tasks done/total |
|---|---|---|
| (no stories) | — | — |

## Bugs

- Open: 0
- Escalated lanes: none

## QA health

| Metric | Value |
|---|---|
| QA verdicts recorded | 0 |
| QA pass rate | — |
| Bug-escape rate | — |
| Mean task wall-time | — |

## Spend

**Runner-reported, advisory, and a FLOOR.** Every figure below is what the agent
client itself said an invocation cost. Plane B has no gateway meter in front of
the client, so nothing corroborates these numbers and nothing bills from them —
they are provenance, not billing truth. They are SPEND, not a cap: neither the
provider account's limit nor any `runners.<role>.budget_usd` this project has
configured. Dispatches on an adapter that cannot report cost are excluded
entirely rather than counted as $0.00, so the true spend is **at least** this.

| Metric | Value |
|---|---|
| Total (all recorded runs) | $0.0000 |
| Priced dispatches | 0 |
| Dispatches that reported nothing | 0 |
| Degradations (adapter cannot price a role) | 0 |

| Story | Spend | Priced dispatches |
|---|---|---|
| (no priced dispatches) | — | — |

## Reference

- Queue: [harness/state/queue.txt](queue.txt)
- Metrics: [harness/state/metrics.jsonl](metrics.jsonl)
- Blockers: [harness/state/blockers.md](blockers.md)
=== scenario: full
# Status — demo

**GENERATED** by `bash harness/engine/core/orchestrator/orchestrate.sh status` (STORY-010 T-CDL-006).
Do not hand-edit — rerun the command to refresh. Hand-written history lives in git.
**As of (newest metrics event):** 2026-10-01T00:00:00+00:00

## Queue

| | Count |
|---|---|
| Done | 0 |
| Pending | 0 |
| Failed (re-queueable) | 1 |
| Skipped (deps/filter) | 0 |

**Next task (priority order):** (queue empty)

## Lanes in flight

Elapsed is measured against the **newest metrics event**, not the clock — the
same instant the "as of" line reports. That is what keeps two `status` runs on
unchanged state byte-identical (STORY-010 AC-3). Spend is the runner-reported
figure for that task, from `runner_cost` events; see the Spend caveat below.

| Lane | Task | Spec | Elapsed | Spend | State |
|---|---|---|---|---|---|
| 1 | T-K1 | docs/k1.md#t-k1 | 31536800s | $1.0000 + 2 unreported | running |
| — | k2 | — | — | no priced dispatch yet | **owner gone** — the next sweep reclaims it |

**Lane pool** (cap 3, max 1/story):

  - lane 2: DISABLED — reason 2
  - lane 3: DISABLED — reason 3

## Tuning numbers

Derived from `metrics.jsonl` alone (`orchestrate.sh lane-metrics`). They
inform a human choosing `max_concurrent_lanes`; the scheduler never tunes
itself.

| Metric | Value |
|---|---|
| Lane utilisation | 50% (10s busy across 2 lane(s) over 20s) |
| Semaphore wait | 3s mean over 2 wait(s), longest 5s |
| Merge-conflict rate | 0% (0 of 4 finished lanes) |

## Stories

| Story | Status | Tasks done/total |
|---|---|---|
| S1 | qa | 1/1 |

## Bugs

- Open: 3 (P0: 1, P3: 1)
- Escalated lanes: STORY-1 STORY-22

## QA health

| Metric | Value |
|---|---|
| QA verdicts recorded | 4 |
| QA pass rate | 75% (3/4) |
| Bug-escape rate | 50% (1/2 passed tasks later grew a bug) |
| Mean task wall-time | 25s over 4 completed tasks |

## Spend

**Runner-reported, advisory, and a FLOOR.** Every figure below is what the agent
client itself said an invocation cost. Plane B has no gateway meter in front of
the client, so nothing corroborates these numbers and nothing bills from them —
they are provenance, not billing truth. They are SPEND, not a cap: neither the
provider account's limit nor any `runners.<role>.budget_usd` this project has
configured. Dispatches on an adapter that cannot report cost are excluded
entirely rather than counted as $0.00, so the true spend is **at least** this.

| Metric | Value |
|---|---|
| Total (all recorded runs) | $3.5000 |
| Priced dispatches | 4 |
| Dispatches that reported nothing | 1 |
| Degradations (adapter cannot price a role) | 2 |

| Story | Spend | Priced dispatches |
|---|---|---|
| S1 | $2.5000 | 3 |
| S2 | $1.0000 | 1 |

## Reference

- Queue: [harness/state/queue.txt](queue.txt)
- Metrics: [harness/state/metrics.jsonl](metrics.jsonl)
- Blockers: [harness/state/blockers.md](blockers.md)
rc=0
--- current-status.md:
# Status — demo

**GENERATED** by `bash harness/engine/core/orchestrator/orchestrate.sh status` (STORY-010 T-CDL-006).
Do not hand-edit — rerun the command to refresh. Hand-written history lives in git.
**As of (newest metrics event):** 2026-10-01T00:00:00+00:00

## Queue

| | Count |
|---|---|
| Done | 0 |
| Pending | 0 |
| Failed (re-queueable) | 1 |
| Skipped (deps/filter) | 0 |

**Next task (priority order):** (queue empty)

## Lanes in flight

Elapsed is measured against the **newest metrics event**, not the clock — the
same instant the "as of" line reports. That is what keeps two `status` runs on
unchanged state byte-identical (STORY-010 AC-3). Spend is the runner-reported
figure for that task, from `runner_cost` events; see the Spend caveat below.

| Lane | Task | Spec | Elapsed | Spend | State |
|---|---|---|---|---|---|
| 1 | T-K1 | docs/k1.md#t-k1 | 31536800s | $1.0000 + 2 unreported | running |
| — | k2 | — | — | no priced dispatch yet | **owner gone** — the next sweep reclaims it |

**Lane pool** (cap 3, max 1/story):

  - lane 2: DISABLED — reason 2
  - lane 3: DISABLED — reason 3

## Tuning numbers

Derived from `metrics.jsonl` alone (`orchestrate.sh lane-metrics`). They
inform a human choosing `max_concurrent_lanes`; the scheduler never tunes
itself.

| Metric | Value |
|---|---|
| Lane utilisation | 50% (10s busy across 2 lane(s) over 20s) |
| Semaphore wait | 3s mean over 2 wait(s), longest 5s |
| Merge-conflict rate | 0% (0 of 4 finished lanes) |

## Stories

| Story | Status | Tasks done/total |
|---|---|---|
| S1 | qa | 1/1 |

## Bugs

- Open: 3 (P0: 1, P3: 1)
- Escalated lanes: STORY-1 STORY-22

## QA health

| Metric | Value |
|---|---|
| QA verdicts recorded | 4 |
| QA pass rate | 75% (3/4) |
| Bug-escape rate | 50% (1/2 passed tasks later grew a bug) |
| Mean task wall-time | 25s over 4 completed tasks |

## Spend

**Runner-reported, advisory, and a FLOOR.** Every figure below is what the agent
client itself said an invocation cost. Plane B has no gateway meter in front of
the client, so nothing corroborates these numbers and nothing bills from them —
they are provenance, not billing truth. They are SPEND, not a cap: neither the
provider account's limit nor any `runners.<role>.budget_usd` this project has
configured. Dispatches on an adapter that cannot report cost are excluded
entirely rather than counted as $0.00, so the true spend is **at least** this.

| Metric | Value |
|---|---|
| Total (all recorded runs) | $3.5000 |
| Priced dispatches | 4 |
| Dispatches that reported nothing | 1 |
| Degradations (adapter cannot price a role) | 2 |

| Story | Spend | Priced dispatches |
|---|---|---|
| S1 | $2.5000 | 3 |
| S2 | $1.0000 | 1 |

## Reference

- Queue: [harness/state/queue.txt](queue.txt)
- Metrics: [harness/state/metrics.jsonl](metrics.jsonl)
- Blockers: [harness/state/blockers.md](blockers.md)
