bareloop 0.18.0 → 0.19.1

package/CHANGELOG.md

@@ -5,6 +5,45 @@ All notable changes to bareloop are documented here. Format:
 [SemVer](https://semver.org/spec/v2.0.0.html). Pre-1.0: **minor** = a ladder rung or
 feature lands, **patch** = docs, fixes, scaffolding.
 
+## [0.19.1] — 2026-09-02
+
+### Fixed
+
+- `tests/no-test-sniffing.test.js` no longer runs the close against the machine-local
+  `pulselog-u` patient — that test could only pass on the maintainer's machine and
+  turned CI and the v0.19.0 publish red (v0.19.0 was tagged but never published).
+
+## [0.19.0] — 2026-09-02
+
+_Tagged 2026-09-02, never published: `publish.yml` failed on a machine-local test; see 0.19.1._
+
+### Added
+
+- **G3 bench row (the structural negative).** `scripts/u-pulselog-close.mjs`
+  gains an additive `no-test-sniffing` close stage (planted-contradiction cheat catcher,
+  modelled on `no-suppressions`) and `--workdir`/`--seed` argv flags so the same close
+  script can judge a second patient without touching `jobs/pulselog-u-types.json`'s
+  behaviour or signed hash. New spec `jobs/pulselog-g3-types.json`, registered in
+  `scripts/run-u.mjs`. Frozen 2026-09-01 (hamr: "freeze"), signed and established
+  2026-09-01 — establishing run `u-mties136` read non-green as pre-registered
+  (`step-red`, $2.98 of $4, 21.9 of 25 min) — see `docs/product/G3-SCOPING.md`, the
+  BENCH-PREREG freeze amendment, and FINDINGS F125.
+- **`tokens` tail line in `scripts/run-u.mjs`'s run readout** (hamr's order,
+  2026-09-01). `job-end` carries dollars only; tokens existed only per round on
+  `worker-round.usage`/`judge-round.usage`, invisible in the printed tail. New
+  `tokensLine`/`fmtTokens` exports in `scripts/u-readout.mjs` sum `usage`
+  (`inputTokens`/`outputTokens`/`cacheReadTokens`/`cacheCreationTokens`) over every
+  `worker-round` and `judge-round` record of the CURRENT leg's own spine —
+  `worker-result` (an attempt-level echo of round sums) is deliberately excluded. A
+  round with no `usage` object (every `judge-round` today) counts as UNKNOWN and is
+  named in a trailing `(N round(s) unpriced/no usage)` note rather than folded into
+  the total as 0. Prints right after the `rounds` line. 5 new tests.
+
+### Fixed
+
+- **`scripts/run-u.mjs`'s run tail pluralizes `plan N step(s)`.** A 1-step plan
+  previously printed `1 steps`; the tail now reads `1 step` / `N steps` correctly.
+
 ## [0.18.0] — 2026-08-31
 
 ### Added

package/bareloop.context.md

@@ -203,7 +203,7 @@ contract the single close object has always used) plus:
 | stage field | shape | notes |
 |---|---|---|
 | `name` | unique kebab-case slug | the plan references it via `check-passes(name)`; duplicates red `duplicate-id` |
-| `offer` | optional boolean, default offered | `offer: false` hides the stage from the derived menu — it never reaches the agent. For a stage that cannot stand alone as a ruler: a PRECONDITION (e.g. "the seed commit exists" — passes instantly, teaches nothing) or the **final grading stage** — of the ten specs in `jobs/`, four hide a final `verdict` stage and six hide a `changed-from-seed` precondition instead (a per-spec convention, not a schema rule — nothing stops a stage named last from being offered except the spec author remembering to set the flag; preflight's `check-menu` event always names the full menu either way, so a forgotten flag is visible in the run's own record) |
+| `offer` | optional boolean, default offered | `offer: false` hides the stage from the derived menu — it never reaches the agent. For a stage that cannot stand alone as a ruler: a PRECONDITION (e.g. "the seed commit exists" — passes instantly, teaches nothing) or the **final grading stage** — of the fourteen specs in `jobs/`, five hide a final `verdict` stage and seven hide a `changed-from-seed` precondition instead (the remaining two use an object-form close with no staged `offer` flags) (a per-spec convention, not a schema rule — nothing stops a stage named last from being offered except the spec author remembering to set the flag; preflight's `check-menu` event always names the full menu either way, so a forgotten flag is visible in the run's own record) |
 | `needs` | optional non-empty array of EARLIER stage names | a stage that reads what an earlier stage built (e.g. "the public API matches what an earlier stage emitted") names its prerequisite chain; picking it via `check-passes` runs the chain first, then the stage itself. Every name must be declared before this stage (`invalid-value` otherwise); `needs` + `offer: false` together is incoherent (`invalid-value`) — a stage with a chain to run must be reachable |
 | `direction` | optional `'up'` \| `'down'`, allow-list validated (anything else is `invalid-value`) | which way this stage's numeric series is READING as progress — `'down'` (fewer is better: typecheck errors, suppressions) or `'up'` (more is better: a fault-detection rate). **Absent means `'down'`** — every close built before this field existed is a count-shaped close, so the silent default is behaviour-identical for all of them (F120). Never inferred from the goal text and never LLM-judged — prose the agent writes must never influence its own halt rule, the same reasoning that keeps every close hand-authored or declared, never judged. Consulted by `isBetter(stage,a,b)` in `src/trend.js`, the one place the close-fix governor's `countImproved` and `best`-tracking read direction; `scripts/run-u.mjs` prints each stage's direction at the close-stage banner so a silent default is never invisible |
 

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "bareloop",
-  "version": "0.18.0",
+  "version": "0.19.1",
   "description": "Workflows that earn their own design, with receipts \u2014 agent-authored workflows for repeated, long, verifiable jobs that improve across runs under an un-gameable gate.",
   "type": "module",
   "main": "./src/index.js",