fieldtest: milestone-42 A/B — promise holds, cold start 9x slower to first family

Milestone-scope consumer run on the release binary at 2f1bace, three scenarios: cold start (help-only) reaches a first single run in 5 commands and a first selecting 4-cell family in 28 with 5 dead ends; the five retired verbs refuse without re-routing; the full retired research scope (grid, walk-forward, monte-carlo bootstrap, generalize) runs document-first over two years of GER40+USDJPY in 2.4s with bit-identical reproduce (21/21). A/B against the sugar-era cycle-300 transcript: comparable to first run, ~9x slower to first family, strictly better artifact — the dominant cost is an introspect --unwired bug on present-but-empty presentation (fixed in-cycle, follow-up commit) plus the missing pre-run coverage view (evidence on the tracker). 4 bugs / 6 friction / 3 spec-gap / 4 working, routed on the tracker. refs #319
2026-07-26 00:21:47 +02:00
parent 2f1baceec7
commit e6b60bf680
20 changed files with 741 additions and 0 deletions
@@ -0,0 +1,280 @@
+# Fieldtest transcript — milestone 42 (sugar retirement, document-first authoring)
+
+Binary: `target/release/aura` built from the working tree at HEAD `2f1bace`
+(`cargo build --release -p aura-cli`, exit 0). Every command below was issued
+through that binary; no implementation source was read.
+
+## Persona
+
+A downstream consumer with no prior aura knowledge, holding only the CLI's own
+help, the scaffolded project's `CLAUDE.md`, and (from scenario 2 on) the public
+README / authoring guide. Scenario 2 additionally carries sugar-era muscle
+memory: `run` / `sweep` / `walkforward` / `mc` / `generalize`.
+
+## Files
+
+| Path | What |
+|---|---|
+| `coldstart-lab/` | scenario 1 — scaffolded by `aura new`, then grown to a 4-cell selecting sweep |
+| `coldstart-lab/documents/cold-process.json` | the process document authored from the `--unwired` ladder |
+| `coldstart-lab/documents/cold-campaign.json` | the campaign document authored from the `--unwired` ladder |
+| `spine-lab/` | scenario 3 — scaffolded, plus an op-script-built tapped blueprint |
+| `spine-lab/blueprints/m42_trend.ops.json` | the op-script (canonical authoring form) |
+| `spine-lab/blueprints/m42_trend.json` | `aura graph build` output, registered as label `m42_trend` |
+| `spine-lab/documents/spine-process.json` | grid → walk_forward → monte_carlo → generalize |
+| `spine-lab/documents/spine-campaign.json` | GER40 + USDJPY, 2 years, 6 axis points |
+| `logs/m42_1_unwired_ladder.sh` + `.txt` | replayable repro of the `--unwired` blind spot |
+| `logs/m42_2_retired_verbs.txt` | the five retired verbs + two retired flags, verbatim |
+| `logs/m42_3_spine.txt` | the spine run, reproduce, chart, tap refusals |
+| `logs/m42_3b_override_family.txt` | the whole `--override` refusal family, both legs |
+
+`runs/` under each lab is gitignored by the scaffolder.
+
+---
+
+## Scenario 1 — COLD START (help + scaffolded docs only)
+
+### To a first successful single run — 4 commands, 0 dead ends
+
+```
+1  $ aura --help
+      → about-text names the two layers, `exec` as the one executor, the
+        document data plane (`aura process` / `aura campaign`), the
+        `introspect --unwired` growth idiom, and `graph introspect --params`
+2  $ aura new --help
+3  $ aura new coldstart-lab
+      created project "coldstart-lab" (data-only; attach native nodes later …)
+4  $ cat coldstart-lab/CLAUDE.md        (scaffolded; 5 bullets, all correct)
+5  $ aura exec blueprints/signal.json
+      {"manifest":{…,"topology_hash":"0d3c6940…"},"metrics":{…,"expectancy_r":1.271…}}
+      EXIT=0
+```
+
+Clean. The scaffolded `CLAUDE.md`'s first bullet is literally the command that
+works.
+
+### To a first multi-cell family with a selection result — 24 commands, 5 dead ends
+
+Discovery (7 commands, all productive):
+
+```
+6  $ aura campaign --help
+7  $ echo '{}' > bare.json; aura campaign introspect --unwired bare.json
+8  $ aura process introspect --unwired bare.json
+9  $ aura process introspect --help
+10 $ aura process introspect --vocabulary        → the 6 stage blocks + meanings
+11 $ aura process introspect --block std::sweep  → metric | select | deflate
+12 $ aura graph introspect --params blueprints/signal.json
+      fast.length:I64 default=2 / slow.length:I64 default=4 / bias.scale:F64 default=0.5
+13 $ aura data list                              → 30+ symbols incl. GER40
+14 $ aura data info GER40
+      {"symbol":"GER40","digits":1,"pipSize":1.0,…}     ← geometry only, no time coverage
+15 $ aura graph register blueprints/signal.json  → 0d3c6940…
+16 $ aura process introspect --metrics           → rankable | gate | generalize tags
+```
+
+Growing the documents, with the refusals verbatim:
+
+```
+17 $ aura process validate documents/cold-process.json    # pipeline: [{"std::sweep":{…}}]
+   aura: process document: a pipeline stage must carry a string "block" id   EXIT=1
+      ← DEAD END 1. Neither `--unwired` ("pipeline (required, list of stage
+        blocks)") nor `--block std::sweep` shows a stage's literal JSON shape.
+        The refusal itself is precise and fixed it in one edit.
+
+18 $ aura process validate  (pipeline: [{"block":"std::sweep",…}])            EXIT=0
+19 $ aura process register documents/cold-process.json  → 26079ba0…
+
+20 $ aura campaign validate documents/cold-campaign.json  # presentation: {}
+   aura: campaign document: missing field `emit`                             EXIT=1
+      ← DEAD END 2. `campaign introspect --unwired` on this same file had
+        reported no open slot for presentation at all (see logs/m42_1_*).
+
+21 $ … "emit": "summary"
+   aura: campaign document: invalid type: string "summary", expected a sequence  EXIT=1
+      ← DEAD END 3 (the refusal names no legal value)
+
+22 $ … "emit": []
+   aura: campaign document: unknown variant `list`, expected one of `I64`, `F64`, `Bool`, `Timestamp`
+      ← DEAD END 4 — the axes `kind` slot; `--unwired` says only
+        "axes (required, map: param name -> { kind, values })".  EXIT=1
+
+23 $ … axes kind → "I64", emit → ["nonsense"]
+   aura: campaign document invalid:
+     presentation.emit: unknown kind "nonsense"                              EXIT=1
+      ← DEAD END 5 (still no enumeration of the legal kinds)
+
+24 $ aura campaign introspect --vocabulary       → std::presentation exists
+25 $ aura campaign introspect --block std::presentation
+     persist_taps  required, list of: equity | exposure | r_equity | net_r_equity
+     emit          required, list of: family_table | selection_report        ← the answer
+26 $ aura campaign validate documents/cold-campaign.json
+     campaign document valid (intrinsic): 1 strategy(ies), 2 axes (4 points), … 1 cell(s)
+     campaign document valid (referential): all references resolve, axes are in the param space
+     campaign document valid (executable): pipeline shape and static guards pass   EXIT=0
+27 $ aura campaign register documents/cold-campaign.json  → dbf47283…
+28 $ aura exec documents/cold-campaign.json
+     4 × {"family_id":"dbf47283-0-GER40-w0-r0-s0-0","report":{…}}
+     {"selection_report":{…,"winner_ordinal":2,"params":[["fast.length",{"I64":3}],
+       ["slow.length",{"I64":8}]],"selection":{"selection_metric":"expectancy_r",
+       "n_trials":4,"raw_winner_metric":0.008618…,"mode":"Argmax"}}}
+     {"campaign_run":{…,"coverage":{"effective_from_ms":1704154499999,
+       "effective_to_ms":1711659420000}}}
+     aura: campaign run 0 recorded: 1 cells                                   EXIT=0
+```
+
+The guessed `from_ms`/`to_ms` (2024-01-01 … 2024-04-01) clipped silently and
+honestly to the archive's real coverage, which the `campaign_run` record then
+reports back as `coverage.effective_*`. There is no *pre*-run way to learn a
+symbol's time coverage: `aura data info` returns geometry only.
+
+**Cold start total: 28 commands, 5 dead ends, all resolved from the surface.**
+
+---
+
+## Scenario 2 — RETURNING SUGAR USER
+
+Full log: `logs/m42_2_retired_verbs.txt`. Condensed:
+
+```
+$ aura run blueprints/m42_trend.json
+error: unrecognized subcommand 'run'
+  tip: a similar subcommand exists: 'runs'                                    EXIT=2
+
+$ aura runs blueprints/m42_trend.json        # following the only tip offered
+error: unrecognized subcommand 'blueprints/m42_trend.json'                    EXIT=2
+
+$ aura sweep / walkforward / mc / generalize <bp>
+error: unrecognized subcommand '<verb>'      # no tip, no retirement notice   EXIT=2
+
+$ aura exec blueprints/m42_trend.json --sweep fast.length=2,3
+error: unexpected argument '--sweep' found
+  tip: to pass '--sweep' as a value, use '-- --sweep'                         EXIT=2
+
+$ aura graph introspect --list-axes blueprints/m42_trend.json
+error: unexpected argument '--list-axes' found                                EXIT=2
+```
+
+Not one of the seven refusals names `exec`, a campaign document, or the fact
+that the verb was retired. Two of them actively point the wrong way (`runs`;
+`-- --sweep`). Re-routing costs exactly one extra command — `aura --help`,
+whose about-text is genuinely excellent — so the honest cost is **2 commands
+to re-orient, 0 commands of it attributable to the refusal**.
+
+---
+
+## Scenario 3 — VALIDATION SPINE (real data, document-first)
+
+Full log: `logs/m42_3_spine.txt`; override family: `logs/m42_3b_override_family.txt`.
+
+Authoring a tapped blueprint in the canonical op-script form worked first try:
+
+```
+$ aura graph build --help          # embeds the full op-list reference, incl. {"op":"tap",…}
+$ aura graph introspect --vocabulary ; --node SMA|Sub|Sign|Bias
+$ aura graph build < blueprints/m42_trend.ops.json > blueprints/m42_trend.json   EXIT=0
+$ aura graph register blueprints/m42_trend.json --name m42_trend
+  registered blueprint b657c714… ; label "m42_trend" -> b657c714…
+$ aura graph introspect --registered
+  m42_trend  b657c714…  fast/slow SMA crossover direction, clamped into the bias contract
+```
+
+`aura graph build` is reachable only from `aura graph --help` / the README —
+the scaffolded project `CLAUDE.md` never mentions how to author a *new*
+blueprint.
+
+The methodology, as a process document (one dead end):
+
+```
+$ aura campaign validate documents/spine-campaign.json     # pipeline had a std::gate after walk_forward
+aura: campaign is not executable:
+  process pipeline is not executable: stage 2: std::gate cannot follow std::walk_forward   EXIT=1
+```
+
+Precise (names the stage index and the illegal adjacency); it does not say what
+*may* follow. Dropping the gate:
+
+```
+$ aura campaign validate documents/spine-campaign.json     # 3 tiers green            EXIT=0
+$ aura campaign register documents/spine-campaign.json  → 0959eda8…
+$ aura exec documents/spine-campaign.json                                              EXIT=0
+  (42 member reports on stdout; 2.4 s wall for 2 instruments × 2 years × 21 WF windows)
+aura: campaign run 0 recorded: 2 cells
+aura: note: tap "net_r_equity" needs a cost model; add a cost block to the campaign document; skipped
+aura: traces persisted: 0959eda8-0 (1 tap(s) x 2 cell(s))
+```
+
+The `campaign_run` record carries the whole retired research scope:
+
+```
+ cell GER40   stages [std::grid, std::walk_forward, std::monte_carlo]
+   pooled_oos mean=-0.004729  p5=-0.0469  p95=0.0381  prob_le_zero=0.59  n=21910
+ cell USDJPY  stages [std::grid, std::walk_forward, std::monte_carlo]
+   pooled_oos mean=-0.033809  p5=-0.0766  p95=0.0177  prob_le_zero=0.885 n=25258
+ generalize   worst_case=-0.251491  sign_agreement=0
+              per_instrument=[GER40 -0.2515, USDJPY -0.0935]  + per-instrument winners
+```
+
+`aura runs family <id> rank expectancy_r` shows every walk-forward window with
+`deflated_score` / `overfit_probability` alongside; `aura reproduce
+0959eda8-0-GER40-w0-r0-s1-0` reports all 21 members **bit-identical**.
+
+`presentation.emit` asked for `family_table` **and** `selection_report`; the run
+emitted 42 family-table lines and **zero** selection-report lines, with no note
+— contrast the `net_r_equity` tap, which announced its own skip.
+
+### The `--override` what-if
+
+```
+$ aura exec documents/spine-campaign.json --override bias.scale=0.25            EXIT=0
+  manifest.params gains ["bias.scale",{"F64":0.25}]; manifest.defaults empties;
+  topology_hash stays b657c714…  (the registered blueprint's id)
+  every R metric bit-identical to the baseline — correct: research runs flat-1R,
+  only the bias *sign* gates firing (invariant 7).
+```
+
+Same operation on the blueprint leg does **not** keep the id:
+
+```
+$ aura exec blueprints/m42_trend.json                          → b657c714…
+$ aura exec blueprints/m42_trend.json --override fast.length=10   → f952a1c8…
+                                       ^ a NO-OP override: 10 is the bound default
+$ aura reproduce f952a1c8…
+aura: no such family 'f952a1c8519a2dbd…'
+```
+
+### The `--override` refusal family (blueprint leg)
+
+```
+--override fast.length            EXIT=2  aura: exec: --override expects NODE.PARAM=VALUE, got `fast.length`
+--override fast.length=notanumber EXIT=2  (same, good)
+--override nosuch.param=1         EXIT=1  aura: axis nosuch.param: names no param of this blueprint (open or bound) — see `aura graph introspect --params <bp>`
+--override fast.length=0          EXIT=1  aura: SMA length must be >= 1                  ← the #319 fix, working
+--override fast.length=-1         EXIT=1  aura: capacity overflow                        ← the same fix, not working
+--override fast.length=-5         EXIT=1  aura: capacity overflow
+--override bias.scale=2           EXIT=1  aura: this blueprint does not compile to a runnable harness: ParamKindMismatch { slot: 0, expected: F64, got: I64 }
+--override bias.scale=2.0         EXIT=0
+--override bias.scale=-5.0        EXIT=1  aura: Bias scale must be > 0
+```
+
+and on the campaign leg for the same inputs:
+
+```
+--override fast.length=7          EXIT=2  aura: exec: --override `fast.length` collides with a declared axis of the campaign; an override overrides a bound value, never an axis
+--override stop_k=1.0             EXIT=1  aura: campaign references do not resolve:
+                                            strategy b657c714…: axis "stop_k" is not in the param space
+--override bias.scale=-5          EXIT=1  aura: campaign references do not resolve:
+                                            strategy b657c714…: axis "bias.scale" declares a kind that is not the param's kind
+```
+
+### Traces
+
+```
+$ aura chart 0959eda8-0                 EXIT=1  aura: no family member has a tap named 'equity'
+$ aura chart 0959eda8-0 --tap r_equity  EXIT=0  (298 KB self-contained HTML on stdout)
+$ aura chart b657c714-GER40-w0          EXIT=1  aura: no recorded run or family 'b657c714-GER40-w0'
+    under runs/traces (check the handle a sweep/walk-forward/campaign run printed for a
+    typo — re-running with this handle as `--trace` will not create it)
+$ aura exec blueprints/signal.json --tap equity=record
+    EXIT=1  aura: the tap plan names 'equity', but the blueprint declares no such tap — declared taps:
+```
@@ -0,0 +1 @@
+/runs
@@ -0,0 +1,4 @@
+# Static project context only (C17); paths only.
+[paths]
+runs = "runs"
+# data = "/path/to/archive"  # the recorded-data root; defaults to the built-in path
@@ -0,0 +1,24 @@
+# coldstart-lab — an aura research project (data-only)
+
+This directory is an aura project: blueprints + research documents over the
+std vocabulary, anchored by `Aura.toml`. There is no crate and no build step.
+
+- Run: `aura exec blueprints/signal.json` (single smoke run, synthetic stream
+  — the starter is closed; all params bound; bound values are defaults —
+  `--override NODE.PARAM=VALUE` may override one for this run (#246))
+- Campaign: `aura exec <campaign.json>` executes a registered campaign
+  document (file or content id) — the multi-cell/axis surface
+- Axes: `aura graph introspect --params blueprints/signal.json` lists the
+  open + bound-overridable axes, RAW `<node>.<param>` names (#328)
+- Native nodes: when the project needs its first project-specific node,
+  `aura nodes new <name>` scaffolds a node crate beside this project and
+  attaches it via `[nodes]` in `Aura.toml` (build it with `cargo build`).
+- Topology is data (`blueprints/*.json`); results land in `runs/`.
+- Execution model: a strategy emits a bias in [-1,+1] per cycle, held as the
+  continuously-tracked target position; a protective stop defines the risk
+  unit R, and quality metrics are R-based. Entry signals become held state
+  via the signal-side latch/edge-pulse idiom (see
+  `aura graph introspect --vocabulary`).
+- Data plane: process/campaign documents are authored directly with
+  `aura process` / `aura campaign`, growing one from a bare `{}` via
+  `aura campaign introspect --unwired`.
@@ -0,0 +1 @@
+{"format_version":1,"blueprint":{"name":"coldstart_lab_signal","doc":"fast/slow SMA difference clamped into a directional bias","nodes":[{"primitive":{"type":"SMA","name":"fast","bound":[{"pos":0,"name":"length","kind":"I64","value":{"I64":2}}]}},{"primitive":{"type":"SMA","name":"slow","bound":[{"pos":0,"name":"length","kind":"I64","value":{"I64":4}}]}},{"primitive":{"type":"Sub"}},{"primitive":{"type":"Bias","name":"bias","bound":[{"pos":0,"name":"scale","kind":"F64","value":{"F64":0.5}}]}}],"edges":[{"from":0,"to":2,"slot":0,"from_field":0},{"from":1,"to":2,"slot":1,"from_field":0},{"from":2,"to":3,"slot":0,"from_field":0}],"input_roles":[{"name":"price","targets":[{"node":0,"slot":0},{"node":1,"slot":0}],"source":"F64"}],"output":[{"node":3,"field":0,"name":"bias"}]}}
@@ -0,0 +1,53 @@
+{
+  "format_version": 1,
+  "kind": "campaign",
+  "name": "cold-start-sweep",
+  "description": "Cold-start 4-cell sweep over the starter blueprint's two lengths on GER40.",
+  "data": {
+    "instruments": [
+      "GER40"
+    ],
+    "windows": [
+      {
+        "from_ms": 1704067200000,
+        "to_ms": 1711929600000
+      }
+    ]
+  },
+  "strategies": [
+    {
+      "ref": {
+        "content_id": "0d3c69401c8e6c9e56661fce348a4c288325a39b7167d7794b30b5141930ebf1"
+      },
+      "axes": {
+        "fast.length": {
+          "kind": "I64",
+          "values": [
+            2,
+            3
+          ]
+        },
+        "slow.length": {
+          "kind": "I64",
+          "values": [
+            8,
+            16
+          ]
+        }
+      }
+    }
+  ],
+  "process": {
+    "ref": {
+      "content_id": "26079ba0317aba8f07740f0fb0616ffed683467618925a3b483ba01d06f74a1f"
+    }
+  },
+  "seed": 0,
+  "presentation": {
+    "persist_taps": [],
+    "emit": [
+      "family_table",
+      "selection_report"
+    ]
+  }
+}
@@ -0,0 +1,9 @@
+{
+  "format_version": 1,
+  "kind": "process",
+  "name": "cold-start-select",
+  "description": "Sweep the axes space and pick the argmax expectancy_r member.",
+  "pipeline": [
+    { "block": "std::sweep", "metric": "expectancy_r", "select": "argmax" }
+  ]
+}
@@ -0,0 +1,33 @@
+#!/bin/sh
+# m42_1 — the --unwired ladder's presentation blind spot, minimal replay.
+# Run from inside any scaffolded project. AURA=<path to the release binary>.
+set -u
+: "${AURA:=aura}"
+t=$(mktemp -d)
+
+echo '{}' > "$t/a.json"
+echo '$ aura campaign introspect --unwired a.json   # a.json = {}'
+"$AURA" campaign introspect --unwired "$t/a.json" | grep -E "^open slot: (data|presentation)"
+
+cat > "$t/b.json" <<'J'
+{"format_version":1,"kind":"campaign","name":"x","seed":0,
+ "data":{"instruments":["GER40"],"windows":[{"from_ms":1,"to_ms":2}]},
+ "strategies":[{"ref":{"content_id":"0"},"axes":{}}],
+ "process":{"ref":{"content_id":"0"}},
+ "presentation":{}}
+J
+echo
+echo '$ aura campaign introspect --unwired b.json   # data filled, presentation left {}'
+"$AURA" campaign introspect --unwired "$t/b.json" | grep -E "^open slot:"
+echo '  ^ only OPTIONAL slots are listed. The two REQUIRED sub-slots of an empty
+    presentation section (persist_taps, emit) are never named — contrast an empty
+    data section, which drills down to instruments/windows (see above).'
+
+echo
+echo '$ aura campaign validate b.json               # the slot the ladder never named'
+"$AURA" campaign validate "$t/b.json" 2>&1 | head -2
+
+echo
+echo '$ aura campaign introspect --block std::presentation   # where the answer actually lives'
+"$AURA" campaign introspect --block std::presentation
+rm -rf "$t"
@@ -0,0 +1,23 @@
+$ aura campaign introspect --unwired a.json   # a.json = {}
+open slot: data (required section: instruments + windows)
+open slot: presentation (required section: persist_taps (equity | exposure | r_equity | net_r_equity) + emit)
+
+$ aura campaign introspect --unwired b.json   # data filled, presentation left {}
+open slot: description (optional, string — a one-line meaning (C29-gated when present))
+open slot: risk (optional, list of stop regimes { vol: { length, k } } | { vol_tf: { period_minutes, length, k } }; absent = one default regime)
+open slot: cost (optional, list of cost components { constant: { cost_per_trade } } | { vol_slippage: { slip_vol_mult } } | { carry: { carry_per_cycle } }; absent = zero costs (net == gross))
+  ^ only OPTIONAL slots are listed. The two REQUIRED sub-slots of an empty
+    presentation section (persist_taps, emit) are never named — contrast an empty
+    data section, which drills down to instruments/windows (see above).
+
+$ aura campaign validate b.json               # the slot the ladder never named
+aura: campaign document: missing field `persist_taps`
+
+$ aura campaign introspect --block std::presentation   # where the answer actually lives
+std::presentation — campaign section: taps to persist + tables to emit (data-level only)
+  persist_taps         required, list of: equity | exposure | r_equity | net_r_equity
+      equity         cumulative pip equity per cycle
+      exposure       signed position exposure per cycle
+      r_equity       cumulative gross R per cycle
+      net_r_equity   cumulative net R per cycle, after the cost model
+  emit                 required, list of: family_table | selection_report
@@ -0,0 +1,69 @@
+# m42_2 — returning sugar user: the five retired verbs, verbatim (release binary @ 2f1bace)
+
+$ aura run blueprints/m42_trend.json
+error: unrecognized subcommand 'run'
+
+  tip: a similar subcommand exists: 'runs'
+
+Usage: aura [OPTIONS] <COMMAND>
+
+For more information, try '--help'.
+EXIT=2
+
+$ aura sweep blueprints/m42_trend.json
+error: unrecognized subcommand 'sweep'
+
+Usage: aura [OPTIONS] <COMMAND>
+
+For more information, try '--help'.
+EXIT=2
+
+$ aura walkforward blueprints/m42_trend.json
+error: unrecognized subcommand 'walkforward'
+
+Usage: aura [OPTIONS] <COMMAND>
+
+For more information, try '--help'.
+EXIT=2
+
+$ aura mc blueprints/m42_trend.json
+error: unrecognized subcommand 'mc'
+
+Usage: aura [OPTIONS] <COMMAND>
+
+For more information, try '--help'.
+EXIT=2
+
+$ aura generalize blueprints/m42_trend.json
+error: unrecognized subcommand 'generalize'
+
+Usage: aura [OPTIONS] <COMMAND>
+
+For more information, try '--help'.
+EXIT=2
+
+$ aura runs blueprints/m42_trend.json      # following the only tip offered
+error: unrecognized subcommand 'blueprints/m42_trend.json'
+
+Usage: aura runs [OPTIONS] <COMMAND>
+
+For more information, try '--help'.
+EXIT=2
+
+$ aura exec blueprints/m42_trend.json --sweep fast.length=2,3   # sugar-era flag muscle memory
+error: unexpected argument '--sweep' found
+
+  tip: to pass '--sweep' as a value, use '-- --sweep'
+
+Usage: aura exec <TARGET>
+
+For more information, try '--help'.
+EXIT=2
+
+$ aura graph introspect --list-axes blueprints/m42_trend.json   # retired discovery flag
+error: unexpected argument '--list-axes' found
+
+Usage: aura graph introspect [OPTIONS]
+
+For more information, try '--help'.
+EXIT=2
@@ -0,0 +1,48 @@
+# m42_3 — the validation spine document-first, GER40 + USDJPY, 2023-01-01..2024-12-31
+# process pipeline: std::grid -> std::walk_forward(90/30/30 rolling, argmax expectancy_r) -> std::monte_carlo(200x5) -> std::generalize
+
+$ aura campaign validate documents/spine-campaign.json
+campaign document valid (intrinsic): 1 strategy(ies), 2 axes (6 points), 2 instrument(s), 1 window(s), 1 regime(s) (default) — 2 cell(s)
+campaign document valid (referential): all references resolve, axes are in the param space
+campaign document valid (executable): pipeline shape and static guards pass
+EXIT=0
+
+$ aura exec documents/spine-campaign.json   # stdout: 42 member reports + campaign_run; stderr below
+EXIT=0
+aura: campaign run 2 recorded: 2 cells
+aura: note: tap "net_r_equity" needs a cost model; add a cost block to the campaign document; skipped
+aura: traces persisted: 0959eda8-2 (1 tap(s) x 2 cell(s))
+stdout lines: 43
+
+# campaign_run record, trimmed to the validation payload:
+ cell GER40 stages ['std::grid', 'std::walk_forward', 'std::monte_carlo']
+   monte_carlo pooled_oos: mean=-0.004729 p5=-0.0469 p95=0.0381 prob_le_zero=0.59 n_trades=21910
+   coverage {'effective_from_ms': 1672618499999, 'effective_to_ms': 1735592220000}
+ cell USDJPY stages ['std::grid', 'std::walk_forward', 'std::monte_carlo']
+   monte_carlo pooled_oos: mean=-0.033809 p5=-0.0766 p95=0.0177 prob_le_zero=0.885 n_trades=25258
+   coverage {'effective_from_ms': 1672617660000, 'effective_to_ms': 1735603200000}
+ generalize: metric=expectancy_r n_instruments=2 worst_case=-0.251491 sign_agreement=0 per_instrument=[['GER40', -0.2514905108214054], ['USDJPY', -0.09353171411968528]]
+ NOTE: presentation.emit asked for family_table AND selection_report;
+       stdout carried 42 family_table lines and ZERO selection_report lines, silently.
+
+$ aura runs families
+{"family_id":"0959eda8-0-GER40-w0-r0-s1-0","kind":"WalkForward","members":21}
+{"family_id":"0959eda8-0-USDJPY-w0-r0-s1-0","kind":"WalkForward","members":21}
+
+$ aura reproduce 0959eda8-0-GER40-w0-r0-s1-0   (first 2 of 21 members)
+0959eda8-0-GER40-w0-r0-s1-0  member fast.length=10, slow.length=100, stop_length=3, stop_k=2  reproduced: bit-identical
+0959eda8-0-GER40-w0-r0-s1-0  member fast.length=10, slow.length=50, stop_length=3, stop_k=2  reproduced: bit-identical
+
+$ aura chart 0959eda8-0                        # default tap is equity; the campaign persisted r_equity
+EXIT=1
+aura: no family member has a tap named 'equity'
+
+$ aura chart b657c714-GER40-w0                 # the literal directory name under runs/traces/0959eda8-0/
+EXIT=1
+aura: no recorded run or family 'b657c714-GER40-w0' under runs/traces (check the handle a sweep/walk-forward/campaign run printed for a typo — re-running with this handle as `--trace` will not create it)
+  ^ names a `--trace` flag that exists on no verb after the retirement.
+
+$ aura exec blueprints/signal.json --tap equity=record   # the scaffolded starter blueprint
+EXIT=1
+aura: the tap plan names 'equity', but the blueprint declares no such tap — declared taps: 
+  ^ the scaffolded starter declares no taps at all; the list after the em dash is empty.
@@ -0,0 +1,61 @@
+# m42_3b — the --override refusal family on both legs (release binary @ 2f1bace)
+# blueprint leg
+
+$ aura exec blueprints/m42_trend.json --override fast.length
+EXIT=2
+aura: exec: --override expects NODE.PARAM=VALUE, got `fast.length`
+
+$ aura exec blueprints/m42_trend.json --override fast.length=notanumber
+EXIT=2
+aura: exec: --override expects NODE.PARAM=VALUE, got `fast.length=notanumber`
+
+$ aura exec blueprints/m42_trend.json --override nosuch.param=1
+EXIT=1
+aura: axis nosuch.param: names no param of this blueprint (open or bound) — see `aura graph introspect --params <bp>`
+
+$ aura exec blueprints/m42_trend.json --override fast.length=0
+EXIT=1
+aura: SMA length must be >= 1
+
+$ aura exec blueprints/m42_trend.json --override fast.length=-1
+EXIT=1
+aura: capacity overflow
+
+$ aura exec blueprints/m42_trend.json --override fast.length=-5
+EXIT=1
+aura: capacity overflow
+
+$ aura exec blueprints/m42_trend.json --override bias.scale=2
+EXIT=1
+aura: this blueprint does not compile to a runnable harness: ParamKindMismatch { slot: 0, expected: F64, got: I64 }
+
+$ aura exec blueprints/m42_trend.json --override bias.scale=-5.0
+EXIT=1
+aura: Bias scale must be > 0
+
+$ aura exec blueprints/m42_trend.json --override bias.scale=0.5
+EXIT=0
+
+# campaign leg
+
+$ aura exec documents/spine-campaign.json --override fast.length=7
+EXIT=2
+aura: exec: --override `fast.length` collides with a declared axis of the campaign; an override overrides a bound value, never an axis
+
+$ aura exec documents/spine-campaign.json --override stop_k=1.0
+EXIT=1
+aura: campaign references do not resolve:
+  strategy b657c7148655fe3774910614031a8dcae6be317f0bbe7de6cbabfeb5ccb1d7cf: axis "stop_k" is not in the param space
+
+$ aura exec documents/spine-campaign.json --override bias.scale=-5
+EXIT=1
+aura: campaign references do not resolve:
+  strategy b657c7148655fe3774910614031a8dcae6be317f0bbe7de6cbabfeb5ccb1d7cf: axis "bias.scale" declares a kind that is not the param's kind
+
+# topology_hash under --override — blueprint leg vs campaign leg
+$ aura exec blueprints/m42_trend.json                          -> topology_hash:
+b657c7148655fe3774910614031a8dcae6be317f0bbe7de6cbabfeb5ccb1d7cf
+$ aura exec blueprints/m42_trend.json --override fast.length=10  (NO-OP: 10 is the bound default) -> topology_hash:
+f952a1c8519a2dbd30fac8facb6149a8774bf4094d552a11e50afecbad9f6ccc
+$ aura reproduce <that hash>
+aura: no such family 'f952a1c8519a2dbd30fac8facb6149a8774bf4094d552a11e50afecbad9f6ccc'
@@ -0,0 +1 @@
+/runs
@@ -0,0 +1,4 @@
+# Static project context only (C17); paths only.
+[paths]
+runs = "runs"
+# data = "/path/to/archive"  # the recorded-data root; defaults to the built-in path
@@ -0,0 +1,24 @@
+# spine-lab — an aura research project (data-only)
+
+This directory is an aura project: blueprints + research documents over the
+std vocabulary, anchored by `Aura.toml`. There is no crate and no build step.
+
+- Run: `aura exec blueprints/signal.json` (single smoke run, synthetic stream
+  — the starter is closed; all params bound; bound values are defaults —
+  `--override NODE.PARAM=VALUE` may override one for this run (#246))
+- Campaign: `aura exec <campaign.json>` executes a registered campaign
+  document (file or content id) — the multi-cell/axis surface
+- Axes: `aura graph introspect --params blueprints/signal.json` lists the
+  open + bound-overridable axes, RAW `<node>.<param>` names (#328)
+- Native nodes: when the project needs its first project-specific node,
+  `aura nodes new <name>` scaffolds a node crate beside this project and
+  attaches it via `[nodes]` in `Aura.toml` (build it with `cargo build`).
+- Topology is data (`blueprints/*.json`); results land in `runs/`.
+- Execution model: a strategy emits a bias in [-1,+1] per cycle, held as the
+  continuously-tracked target position; a protective stop defines the risk
+  unit R, and quality metrics are R-based. Entry signals become held state
+  via the signal-side latch/edge-pulse idiom (see
+  `aura graph introspect --vocabulary`).
+- Data plane: process/campaign documents are authored directly with
+  `aura process` / `aura campaign`, growing one from a bare `{}` via
+  `aura campaign introspect --unwired`.
@@ -0,0 +1 @@
+{"format_version":1,"blueprint":{"name":"m42_trend","doc":"fast/slow SMA crossover direction, clamped into the bias contract","nodes":[{"primitive":{"type":"SMA","name":"fast","bound":[{"pos":0,"name":"length","kind":"I64","value":{"I64":10}}]}},{"primitive":{"type":"SMA","name":"slow","bound":[{"pos":0,"name":"length","kind":"I64","value":{"I64":50}}]}},{"primitive":{"type":"Sub","name":"spread"}},{"primitive":{"type":"Sign","name":"dir"}},{"primitive":{"type":"Bias","name":"bias","bound":[{"pos":0,"name":"scale","kind":"F64","value":{"F64":1.0}}]}}],"edges":[{"from":0,"to":2,"slot":0,"from_field":0},{"from":1,"to":2,"slot":1,"from_field":0},{"from":2,"to":3,"slot":0,"from_field":0},{"from":3,"to":4,"slot":0,"from_field":0}],"input_roles":[{"name":"price","targets":[{"node":0,"slot":0},{"node":1,"slot":0}],"source":"F64"}],"output":[{"node":4,"field":0,"name":"bias"}],"taps":[{"name":"spread","from":{"node":2,"field":0}}]}}
@@ -0,0 +1,17 @@
+[
+  {"op":"name","name":"m42_trend"},
+  {"op":"doc","text":"fast/slow SMA crossover direction, clamped into the bias contract"},
+  {"op":"source","role":"price","kind":"F64"},
+  {"op":"add","type":"SMA","name":"fast","bind":{"length":{"I64":10}}},
+  {"op":"add","type":"SMA","name":"slow","bind":{"length":{"I64":50}}},
+  {"op":"add","type":"Sub","name":"spread"},
+  {"op":"add","type":"Sign","name":"dir"},
+  {"op":"add","type":"Bias","name":"bias","bind":{"scale":{"F64":1.0}}},
+  {"op":"feed","role":"price","into":["fast.series","slow.series"]},
+  {"op":"connect","from":"fast.value","to":"spread.lhs"},
+  {"op":"connect","from":"slow.value","to":"spread.rhs"},
+  {"op":"connect","from":"spread.value","to":"dir.value"},
+  {"op":"connect","from":"dir.value","to":"bias.signal"},
+  {"op":"expose","from":"bias.bias","as":"bias"},
+  {"op":"tap","from":"spread.value","as":"spread"}
+]
@@ -0,0 +1 @@
+{"format_version":1,"blueprint":{"name":"spine_lab_signal","doc":"fast/slow SMA difference clamped into a directional bias","nodes":[{"primitive":{"type":"SMA","name":"fast","bound":[{"pos":0,"name":"length","kind":"I64","value":{"I64":2}}]}},{"primitive":{"type":"SMA","name":"slow","bound":[{"pos":0,"name":"length","kind":"I64","value":{"I64":4}}]}},{"primitive":{"type":"Sub"}},{"primitive":{"type":"Bias","name":"bias","bound":[{"pos":0,"name":"scale","kind":"F64","value":{"F64":0.5}}]}}],"edges":[{"from":0,"to":2,"slot":0,"from_field":0},{"from":1,"to":2,"slot":1,"from_field":0},{"from":2,"to":3,"slot":0,"from_field":0}],"input_roles":[{"name":"price","targets":[{"node":0,"slot":0},{"node":1,"slot":0}],"source":"F64"}],"output":[{"node":3,"field":0,"name":"bias"}]}}
@@ -0,0 +1,58 @@
+{
+  "format_version": 1,
+  "kind": "campaign",
+  "name": "m42-spine-ger40-usdjpy",
+  "description": "Walk-forward + bootstrap + cross-instrument floor for the m42 trend blueprint on two instruments.",
+  "data": {
+    "instruments": [
+      "GER40",
+      "USDJPY"
+    ],
+    "windows": [
+      {
+        "from_ms": 1672531200000,
+        "to_ms": 1735603200000
+      }
+    ]
+  },
+  "strategies": [
+    {
+      "ref": {
+        "content_id": "b657c7148655fe3774910614031a8dcae6be317f0bbe7de6cbabfeb5ccb1d7cf"
+      },
+      "axes": {
+        "fast.length": {
+          "kind": "I64",
+          "values": [
+            5,
+            10,
+            20
+          ]
+        },
+        "slow.length": {
+          "kind": "I64",
+          "values": [
+            50,
+            100
+          ]
+        }
+      }
+    }
+  ],
+  "process": {
+    "ref": {
+      "content_id": "14e9ce32b1ea050c23e66df99b182e0de0c3634063c9a887407c98852b75fbae"
+    }
+  },
+  "seed": 7,
+  "presentation": {
+    "persist_taps": [
+      "r_equity",
+      "net_r_equity"
+    ],
+    "emit": [
+      "family_table",
+      "selection_report"
+    ]
+  }
+}
@@ -0,0 +1,29 @@
+{
+  "format_version": 1,
+  "kind": "process",
+  "name": "m42-validation-spine",
+  "description": "Grid to walk-forward, bootstrap the pooled OOS R, then floor across instruments.",
+  "pipeline": [
+    {
+      "block": "std::grid"
+    },
+    {
+      "block": "std::walk_forward",
+      "in_sample_ms": 7776000000,
+      "out_of_sample_ms": 2592000000,
+      "step_ms": 2592000000,
+      "mode": "rolling",
+      "metric": "expectancy_r",
+      "select": "argmax"
+    },
+    {
+      "block": "std::monte_carlo",
+      "resamples": 200,
+      "block_len": 5
+    },
+    {
+      "block": "std::generalize",
+      "metric": "expectancy_r"
+    }
+  ]
+}
				`@@ -0,0 +1 @@`
				{"format_version":1,"blueprint":{"name":"coldstart_lab_signal","doc":"fast/slow SMA difference clamped into a directional bias","nodes":[{"primitive":{"type":"SMA","name":"fast","bound":[{"pos":0,"name":"length","kind":"I64","value":{"I64":2}}]}},{"primitive":{"type":"SMA","name":"slow","bound":[{"pos":0,"name":"length","kind":"I64","value":{"I64":4}}]}},{"primitive":{"type":"Sub"}},{"primitive":{"type":"Bias","name":"bias","bound":[{"pos":0,"name":"scale","kind":"F64","value":{"F64":0.5}}]}}],"edges":[{"from":0,"to":2,"slot":0,"from_field":0},{"from":1,"to":2,"slot":1,"from_field":0},{"from":2,"to":3,"slot":0,"from_field":0}],"input_roles":[{"name":"price","targets":[{"node":0,"slot":0},{"node":1,"slot":0}],"source":"F64"}],"output":[{"node":3,"field":0,"name":"bias"}]}}