bareloop 0.56.0 → 0.58.0

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (93) hide show
  1. package/CHANGELOG.md +80 -0
  2. package/README.md +3 -3
  3. package/bareloop.context.md +75 -214
  4. package/package.json +2 -2
  5. package/src/authorflow.js +71 -136
  6. package/src/authoring.js +5 -124
  7. package/src/authorjob.js +125 -417
  8. package/src/authorreadout.js +105 -355
  9. package/src/authorrun.js +70 -234
  10. package/src/authorscout.js +9 -1
  11. package/src/blockasks.js +12 -17
  12. package/src/bundlerun.js +4 -1
  13. package/src/checkblock.js +297 -75
  14. package/src/checklist.js +1 -1
  15. package/src/checkpick.js +9 -5
  16. package/src/cli.js +3 -3
  17. package/src/declaredclose.js +45 -112
  18. package/src/destpoint.js +1 -1
  19. package/src/draftspend.js +20 -4
  20. package/src/fencepointcall.js +76 -0
  21. package/src/guarddescriptions.js +14 -0
  22. package/src/index.js +3 -43
  23. package/src/interviewrun.js +13 -8
  24. package/src/job.js +20 -1
  25. package/src/jobblock.js +11 -46
  26. package/src/judged.js +13 -1229
  27. package/src/kinds.js +10 -348
  28. package/src/noteask.js +124 -0
  29. package/src/pagedry.js +27 -5
  30. package/src/pagemap.js +7 -29
  31. package/src/pagerules.js +18 -3
  32. package/src/panel/authorroutes.js +15 -4
  33. package/src/panel/authorsession.js +201 -167
  34. package/src/panel/importrun.js +0 -1
  35. package/src/panel/index.html +43 -18
  36. package/src/panel/jobplan.js +54 -19
  37. package/src/panel/server.js +90 -69
  38. package/src/panel/status.js +3 -0
  39. package/src/pickrecord.js +15 -54
  40. package/src/planrun.js +4 -48
  41. package/src/promptregisters.js +7 -17
  42. package/src/providerrows.js +0 -36
  43. package/src/providers.js +4 -25
  44. package/src/questionbox.js +34 -30
  45. package/src/reviewdoor.js +1 -1
  46. package/src/rowresults.js +69 -0
  47. package/src/rubricstep.js +136 -28
  48. package/src/run.js +2 -15
  49. package/src/stagewords.js +24 -0
  50. package/src/userrun.js +14 -111
  51. package/types/authorflow.d.ts +51 -137
  52. package/types/authoring.d.ts +1 -14
  53. package/types/authorjob.d.ts +14 -55
  54. package/types/authorreadout.d.ts +52 -80
  55. package/types/authorrun.d.ts +20 -0
  56. package/types/blockasks.d.ts +5 -11
  57. package/types/checkblock.d.ts +147 -28
  58. package/types/checklist.d.ts +2 -2
  59. package/types/checkpick.d.ts +5 -3
  60. package/types/cli.d.ts +1 -3
  61. package/types/declaredclose.d.ts +5 -33
  62. package/types/draftspend.d.ts +16 -2
  63. package/types/fencepointcall.d.ts +68 -0
  64. package/types/guarddescriptions.d.ts +7 -0
  65. package/types/index.d.ts +2 -4
  66. package/types/job.d.ts +8 -0
  67. package/types/jobblock.d.ts +8 -38
  68. package/types/judged.d.ts +7 -357
  69. package/types/kinds.d.ts +4 -104
  70. package/types/noteask.d.ts +76 -0
  71. package/types/pagedry.d.ts +13 -0
  72. package/types/pagemap.d.ts +2 -5
  73. package/types/pagerules.d.ts +7 -0
  74. package/types/panel/authorsession.d.ts +37 -5
  75. package/types/panel/jobplan.d.ts +15 -3
  76. package/types/panel/server.d.ts +46 -13
  77. package/types/panel/status.d.ts +5 -0
  78. package/types/pickrecord.d.ts +8 -13
  79. package/types/planrun.d.ts +1 -22
  80. package/types/providerrows.d.ts +0 -31
  81. package/types/providers.d.ts +3 -19
  82. package/types/questionbox.d.ts +20 -6
  83. package/types/rowresults.d.ts +23 -0
  84. package/types/rubricstep.d.ts +43 -7
  85. package/types/run.d.ts +1 -20
  86. package/types/stagewords.d.ts +11 -0
  87. package/types/userrun.d.ts +0 -1
  88. package/src/calibrate.js +0 -702
  89. package/src/cardauthor.js +0 -502
  90. package/src/fitcheck.js +0 -168
  91. package/types/calibrate.d.ts +0 -201
  92. package/types/cardauthor.d.ts +0 -358
  93. package/types/fitcheck.d.ts +0 -119
package/CHANGELOG.md CHANGED
@@ -5,6 +5,86 @@ All notable changes to bareloop are documented here. Format:
5
5
  [SemVer](https://semver.org/spec/v2.0.0.html). Pre-1.0: **minor** = a ladder rung or
6
6
  feature lands, **patch** = docs, fixes, scaffolding.
7
7
 
8
+ ## [0.58.0] - 2026-11-16
9
+
10
+ ### Added
11
+
12
+ - **Draft the plan button.** The plan menu's first button reads "Draft the plan" (it only starts drafting, never signs); its footer and the CLI Confirm line say the same.
13
+ - **YOUR ANSWERS in the draft readout and the Audit drafting rows say what was decided** ("agreed", "you wrote", "dropped"), never the question that quotes a replaced rule's old words (found by the full test gate; walked live 2026-11-16). One owner, `decidedWords` in `src/draftspend.js`; older logs with no words keep "you were asked / you said".
14
+
15
+ ### Changed
16
+
17
+ - **The author CLI's "SIGNING PREPARED" block prints the same draft readout (hamr 2026-11-15).** `buildReadout` from the signed spec, the metered call count and cost, and the spec hash; the text the panel Chat shows. Pinned from source (the runner is reachable only after paid calls); not run end to end.
18
+ - **The panel Chat shows the draft readout above the signed-ready plan card (hamr 2026-11-15, copied from fwdloop v0.34.0).** The final plan message carries `readout: {text, heads}` built by `buildReadout` from the signed spec, the draft log and the session's figures (rounds = draft-log calls, cost with "at least" when a call is unpriced, hash, your answers), redrawn on Drop with the new hash (never twice); the page draws it as wrapped text with bold heads. Rubric and repo jobs. Display only, not in the hash, no new store. Rendered check: see the commit notes.
19
+ - **The draft readout's one owner (`buildReadout`, `READOUT_HEADS`, `readoutHeadIndexes` in `src/authorreadout.js`; hamr 2026-11-15, copied from fwdloop v0.34.0).** Pure plain text of what the person is about to sign, built from the signed spec through the card's own readers (`cardOfClose` rows and `[you]/[agent]/[default]` tags, `fenceRowText`, `refusedOf`), masked by `redactSecrets`; display only, never signed or stored.
20
+ - **Regression test for Add/Drop on a Suggested phrase through the panel's `rowEdit` (test only).** `tests/checklist-door.test.js`: before Add the phrase is Suggested with the cited default and the count Not checked (true reason); Add on the phrase gives a facts line and offers the count; Add on the count builds it; Drop returns to the first state.
21
+ - **A result count's Not-checked reason is true when its fact is only Suggested (live run `mwi76vh8`).** The Audit said "a result count needs a fact to count; this line has none" while the line's fact sat under Suggested. With a Suggested fact on the line the reason now reads "a result count needs a fact to count; add the suggested fact on this line to count its results" (`REASON.countNeedsSuggested`, picked by code); a line with no fact keeps the old words.
22
+ - **The Audit Checks header's word "checks" matches the row result words (hamr 2026-11-15).** It now uses the `.st-word` look at the same 12px as the "passed" below it. Display only.
23
+ - **The confirm card lists a look-dropped fact once (live run `mwi76vh8`).** A dropped place fact with no `~` rule ("Boston place") showed twice before Draft the plan: the look's own line and the check block's `! [agent]` row ("boston place"). The look's line is now left out when the block already has the row. Display only.
24
+ - **A look's phrase the person's own `~` rule does not name is no check of that rule (rubric jobs; hamr 2026-11-15, live run `mwi76vh8`).** The look tied the page phrase "Select flight" to the rule "be friendly" and it was signed as a `~ [you]` phrase check. A phrase or place whose words the rule's own words do not contain (case-blind) is now an ordinary `! [agent]` row, placed (job-derived with Drop, or Suggested with Add) by the existing split rule, and the `~` rule is Not checked. A phrase the rule names keeps its link on the rule's row; a version or a date keeps the place-rule path; money, time, number and date links are unchanged. A name the AI gives a phrase or place fact is no proof either: such a fact is job-derived (built, Drop) only when its own words are in the job line or a `~` rule, else it is Suggested (Add).
25
+ - **A READING question keeps a text box (hamr 2026-11-15 ruling).** It offers Agree, Drop it and "Or type the right reading:". Drop it leaves the rule Not checked with `REASON.left`. A typed correction REPLACES the rule (one in, one out) and code proves it as it proves any rule; words that carry no provable bound go to Not checked with `REASON.noRule`, never a second question. A proposed-check question stays Agree / Drop it with no box, a rule nothing can check stays unasked, and typing stays retired for everything else.
26
+ - **A check question is buttons only (rubric and repo jobs; hamr 2026-11-15).** The question box for a look's reading or a proposed check has Agree (and Drop it on a proposed check) and no text box to add a comment or own words; the chat's typing box ("answer here when the draft asks you") is no longer drawn while nothing asks to be typed (no ask, the plan menu, the final card, beside a question box), and still appears for the asks that need typing (language, goal). The typed own-words answer is retired: the check block's `rule` choice, the `or type your own words` / `or type how to check it` wording of the questions, `stripAgreeLead` and `HOW_TO_ANSWER` are gone; `send` refuses typed words for a check question. A rule is now changed only by Revise. The drafter's own try-3 questions, the Destination question and the CLI keep their typed answers (they have no buttons).
27
+ - **A rule no check can prove is not asked: it is Not checked at once (rubric and repo jobs; hamr 2026-11-15, option A).** Before, the person was asked "I can't check this. Drop it?" with Drop it as the only answer. Now the check block marks such a rule Not checked with the code reason `REASON.noRule` ("No rule can check this, so you judge it at Accept."), and the draft goes on. Questions that offer Agree (a look's reading, a proposed count check) stay. Unasked rules do not count toward the 3-question stop (`MAX_QUESTIONS`): only asked questions are counted. The `cant-check` / `no-check` asks, `REASON.cantCheck` / `cantCheckRepo` and `CANT_CHECK_WORDS` are gone.
28
+ - **Revise shows on the final card (rubric and repo jobs).** The button was hidden once the draft reached the signed-ready card (phase `prepared`, no pending ask), because `reviseShownFor` looked only for a pending plan menu; to change the job the person had to clear everything and start over. It now shows there too, with the same 2-revises-then-Start-over count. Seen live 2026-11-15 (run `mwhvvvap`).
29
+ - **The pointing call is shown the folders that hold outside files (`test/`) beside the files.** The live draft `smwi20ir86wxj` showed only files, so a rule "don't edit test/" could not be pointed at its folder. The proof is unchanged.
30
+ - **The pointing call's reply and each edit's accept or refusal reason are in the draft log** (`kind: "fence-point"`: `edits`, `results`, `reason`, secret-masked). Before, the live draft `smwi20ir86wxj` showed only that the call ran.
31
+ - **A drafter line refusal is ignored when a stage or a guard's `guardLines` covers the line** (the rubric's rule: only a line no check could be built for is refused). The live draft `smwi20ir86wxj` showed an extra `No rule can check this` row on a line with three `!` stage checks.
32
+ - **A repo `~` rule is kept by the write fence only when the AI points at it and code proves it (the rubric's order: point, code proves, asks, draft).** One small call BEFORE the asks (`src/fencepointcall.js`, thinking off, one retry, no JSON repair, booked in the draft log) shows the AI each `~` rule's own words and the repo paths outside the write scope, and gets row edits back in the rubric's shape (`{"edits":[{"op":"fence","line","rule","path"}]}`); they go through `CheckBlock.applyDelta` like every rubric edit. Code proves each (the path is on the listing, the rule's own words name it, it lies outside the write scope). Not pointed, or the proof fails: the same question as a rubric rule. Code alone never keeps a rule (so "the tests in test/ must pass" is asked). Called only when the job has `~` rules and the listing holds paths outside the fence; a failed call never stops the draft. The drafter's declaration carries no pointing field.
33
+ - **A line the drafter refused is Not checked with code's reason.** The drafter's refusal is only the signal that a line has no check; the reason on the signed row and in `closeDecl.refused` is `REASON.noRule` (the one table), so the drafter's free text is no second reason owner.
34
+ - **The repo-only row builder is gone.** `src/repolist.js` (`buildRepoRows`, `proveFenceClaim`, `NO_CHECK_WORDS`, its "AI said" reason strings) is deleted: the repo job's rows, reasons (`REASON` in `src/checkblock.js`, the one table) and proofs are the check block's. The view readers keep reading the signed rows of both check types (`!` stage rows and fence-kept `~` rows), so no view changed.
35
+ - **A repo `! [agent]` work stage is a droppable row of the check block, and a Drop is carried by Revise.** In the block a stage row can be Dropped and Added back (never the last check on a line it serves, never the last work stage); the signed list signs it `dropped` and the close runs without it. A Drop made on the final card (`dropStage`) is written to the same record a rubric Drop writes (`choiceEntries`, `src/checkblock.js`), so Revise keeps it; code still judges it on the new draft.
36
+ - **The repo drafter no longer writes `fenceKept` or a rule-level `refused`.** Both are gone from the declaration schema (a `refused` item is a whole job line: `{line, reason}`), from the drafter prompt (the fence-claim paragraph and its legal example), from the draft validator and from `closeDecl` (`fenceKept` leaves `CLOSE_DECL_FIELDS`; it was unreleased). Code decides what the write fence keeps and builds `refused` from the rows. A drafter that still writes either is ignored. An older signed spec that carries `fenceKept` still validates, displays and Reuses (only its shape is read).
37
+ - **A repo job runs through the SAME check block as a rubric job (hamr 2026-11-15: "same flow as rubric, no exceptions; only the TYPE of checks varies").** Its `~` rules, its `!` work stages and its `$` rows are rows of one `CheckBlock` (`src/checkblock.js`, `repo` option); `signedRows`, `closeView`, `setGuards`, the ask rounds (`askRounds` in `src/authorjob.js`) and the Drop / Revise-carry doors are the rubric's own. A `~` rule is kept by the write fence ONLY when code finds a repo path (file, or a folder written `folder/`) in the person's own rule words that lies outside the write scope; the drafter's `fenceKept` claim is no longer read (the live run `mwhvvvap` kept `no new dependencies at all` via package.json and `keep the comment at the top of src/sum.js` via a tracked `.bareloop/` copy). Any other rule becomes a question in the question box (Agree is not offered; Drop it leaves it Not checked, own words replace the rule and are proved again), and a repo question never says "on the page". `refused` is built from the rows in code, so a rule is no longer shown both kept and Not checked.
38
+ - **The files listing the drafter and code's proofs read no longer lists bareloop's own stores.** A patient that tracks `.bareloop/`, `.smoke/` or `.litectx/` files (the live patient tracks 26) no longer offers them as repo files, so a leftover `.bareloop/wt/.../src/sum.js` cannot "keep" a rule. Root prefixes only; `src/.smoke/x` stays the person's own path (`src/authorscout.js` `seedFileList`).
39
+ - **The `$ [default]` guard and fence rows read a short plain question, not a raw guard name plus a long sentence.** `did the run change a file?`, `no checker silenced (@ts-ignore, eslint-disable, # type: ignore)?`, `only files in src/ changed?` (the fence's words are built from the signed fence). One owner, `src/stagewords.js` (`guardQuestionText`, `fenceQuestionText`); the plan card, Job tab, Audit, Result tab and the CLI plan print all read it. An older close with no list keeps its old guard lines.
40
+ - **One list for every rubric check (hamr 2026-11-14, Addendum (c); docs/product/ITEM33-BUILD.md).** Every check a person sees on a rubric job is a row of the one list (`~ [you]`, `! [agent]`, `$ [default]`), code decides, and a view shows nothing that is not a row. Built in six pieces:
41
+ - piece 1: the `$` floor row (the line's file must exist, be complete JSON, and carry text) is drawn on every line on the card, the Job tab and the Audit; a facts-only line says its floor has no text check. The separate "line N file" note of the protections folds into it (an older close with no floor row keeps the note).
42
+ - piece 2: the mandatory guards (`changed-from-seed`, and `no-suppressions` where a class has it) are `$ [default]` rows of the signed list, ONCE for the whole job (`closeDecl.list`, row `kind: 'guard'`, line 0, key `J.$guard.<name>`); the close validator admits them, `listMatchesClose` checks each names a guard stage the close runs, and "Always checked" keeps only the write fence for a job that signs them. `GUARD_DESCRIPTIONS` moved to `src/guarddescriptions.js` (still exported from `authorflow.js`).
43
+ - piece 3: every check item the card and the Job tab draw carries the key of its built list row (a pick, the cited default, the numbers tie and the note too). The CLI plan print (`confirmPlanLines`) draws each check with its line and row tag. A close that carries the list with its guard rows draws no model free-text `notes`, and its Job `success` / `guardrails` fields are the rows and the write fence.
44
+ - piece 4: each row gets its OWN pass or fail. `gradeStep` hands out bareguard's raw gaps; `src/rowresults.js` maps each gap to the one row that owns the check it names; `runDeclaredStages` (given the signed list by `planrun` and the review door) records `rows` and `unclaimed` on each stage entry of the close; the Audit and the Result tab show a mark per row and the reason of a failed one. A gap that names no single row (a quote-level gap of a fact part, a part with no results) is never spread over rows: it is shown as its own failed item of the line.
45
+ - piece 5: the try-3 question replaces. The drafter tags each question with `rule` (a `~` rule the line really has) or `new`; `normalizeQuestions` drops a rule the line does not have, a question with no tag and one with both, with the reason. An answer about a rule REPLACES that rule (one in, one out), `new` adds one; the ask shows "This replaces: <old words>" or "This adds a new rule to line N"; each answered question is logged as draft-log kind `try3-question`.
46
+ - prose row (hamr's ruling Q2): a line that writes text gets a signed `prose` row (`kind: 'prose'`, Not checked, `$ [default]`, "what the text says — you judge it at Accept"), built by the check block; every view draws the line from that row, and the stage-derived `proseNotCheckedItems` feeds nothing for a list-built close. It is never a check (`hasBuiltCheck` skips it).
47
+ - fence row (hamr 2026-11-15): the write fence is a `$ [default]` row of the list, once for the whole job (`kind: 'fence'`, line 0, key `J.$fence`, signed like the guard rows; its words come from the fence the close signs), drawn under "other checks" on the confirm card, the signed card, the Job tab and the Audit (where it reads the gate's own record: held · N writes blocked) and in the CLI print. "Always checked" is gone for a job that signs it; an older close keeps it. The fence is not a check (`hasBuiltCheck` never counted it: it is not a row of the block).
48
+ - a `$` default that does not apply to a line kind is not a Not checked item: a facts line's cited default ("this line finds facts, so it has no text to quote") is no longer drawn (the floor row says the line has no text check); it stays in the list, and stays drawn on a text line that cannot quote (several inputs).
49
+ - tags (hamr 2026-11-15): every row label reads `~ [you]`, `! [agent]`, `$ [default]` on the page (`byTagOf`) and in the CLI print.
50
+ - A signed spec or old run is never repaired: a close with no list, or a list with no guard row, keeps today's display and the same verdicts.
51
+ - **Repo jobs join the one list (hamr 2026-11-15, Addendum (d); docs/product/ITEM33-BUILD.md).** A repo job signs `closeDecl.list`: `~ [you]` one row per rule, `! [agent]` one row per AI work stage, `$ [default]` the guards and the write fence once. hamr's law: a rule shows as kept only when code proves it; everything else is Not checked. Every view reads the rows; an old signed repo spec keeps today's display. (The rows are now built by the one check block, see above.)
52
+
53
+ ## [0.57.0] - 2026-11-14
54
+
55
+ ### Removed
56
+
57
+ - **BREAKING: the old judge's runtime engine is gone (hamr 2026-11-14: "i don't need old runs in anything").** `src/judged.js` keeps only `resolveJudge` / `resolveJobJudge` (the judge identity the spec stamp and the workflow key read); the locate call, the owned rulebook, `decide`, the calibration set's shape and `defaultJudgeLoop` are deleted. The `judged-floor` kind is no longer a live kind or in `KIND_CATALOGUE` (`RETIRED_KINDS` and `judged-floor-gone` go with it), `closeDecl.calibration` is an unknown field, and `runJob` / `runPlan` / `cliMain` no longer take `judgeProvider`, `judgeModel` or `judgeRates` (no `JUDGE_API_KEY` is read; `buildRunnerProviders` returns `{provider, providerFor}`). Library exports removed: `JUDGE_RULES`, `JUDGE_RULE_IDS`, `runLocate`, `decide`, `CALIBRATION_SIZE`, `validateCalibrationSet`, `expectedOf`, `JUDGE_ATTEMPTS`, `MAX_JUDGED_PATHS`. The panel Summary box no longer shows a `judge:` line. A signed spec with a `judged-floor` stage or PASS/FAIL lines now also refuses at `run-u --spec` in the plain words ("draft it fresh"), as it does at Reuse, at a scheduled fire and at the bundle door. Kept: the learnings ledger and how it reads old rows, and the spend readers of an old spine's `judge-round` records.
58
+ - **BREAKING: PASS/FAIL judge examples are gone (hamr 2026-11-14, rubric vs deterministic split, piece 1).** A `~ PASS:` / `~ FAIL:` line is refused at $0 in plain words ("PASS/FAIL examples are gone — write what you want as a ~ rule."), on the panel and the CLI. `jobblock.js` no longer exports `renderExamples` or `jobBlockFitsCheck`, a parsed job line has no `examples` field, the soft-green question set is The job alone (no Judge Examples answer), and the hint text and the panel's rubric placeholder no longer mention the examples. The `~` / `!` / `$` check list (with Drop and Add) covers what the examples did.
59
+ - **BREAKING: the old judge is gone (hamr 2026-11-14, rubric vs deterministic split, piece 2).** `src/cardauthor.js` (the judge-card compile) and `src/calibrate.js` (the 10-case calibration and its signing gate, gate 4 of `prepareSigning`) are deleted, with their library exports: `proposeJudgedArtifacts`, `signJudgedArtifacts`, `foldJudgedArtifacts`, `runCalibration`, `INJECTION_LOCATE_BATTERY`, `CASUALTY_AXES`, and `authorjob`'s `seedProofIsCalibration`. `prepareSigning` takes no `judgeLoop`/`judgeModel`/`onJudgeCost`/`calibrateFn`/`ceilingUsd`/`priorCalls` and returns no `gates.calibration`; `authorCloseForJob` takes no `judgeModel`/`signerFix`/`proposeFn`/`compileOpts` and returns no `judged`. `judged-floor` is no longer on any composer menu (`classMenu`), and a composed or signed close that still carries one is refused at $0 as `judged-floor-gone` ("The old judge is gone. Write what you want checked as a ~ rule on the check list, or draft the job fresh."). The panel and the CLI author run lose the `calibrating` step, the calibration readout and the judge-key checks. The `judged-floor` runtime engine is removed (see the BREAKING bullet above): an old signed spec with a `judged-floor` stage is refused in plain words, not run.
60
+ - **BREAKING: the fit check is gone (hamr 2026-11-14, piece 3).** `src/fitcheck.js` (v0.45.0, "I found N of M of your asks") is deleted: no paid fit call before the look, no Go on / Stop question, no `fit-stopped` stop, no `fit-check` drafting-log row; `authorCloseForJob` takes no `fitFn`/`fitAsk`/`onFit` and the panel session no `fitFn`. One place for all checks: the check list. "A run needs at least one check" stays, from the check list.
61
+
62
+ ### Fixed
63
+
64
+ - check list: a `!` phrase carrying a source (ledger-suggest adds) was judged against every look fact with no source or the same source, on any line, because the money filter in `matchBang` had `&&` and `||` ungrouped; a phrase on a line with no money fact was wrongly Not checked "in none of its results" when another line had one. The filter now reads money facts on this line only, then the source test. Found by /branch-review on this branch; `tests/checkblock-bang-result.test.js`.
65
+ - check list: one rule, one Not-checked row on the confirm card (live run mwgsl7zx: after Revise the person's earlier Drop is kept and the look dropped the same `~` rule again, so the card listed it twice, "the page never shows a time right after ..." and "You chose to leave it Not checked ..."). A `~` rule the check list says Not checked (`CheckBlock.saysRule`) is said by its own row alone, with the person's choice as its reason; the look's line for that rule is no longer drawn beside it (`accepted.notChecked`, the drafter's list, is unchanged). The block itself already held one row (signed list and card); the second row was the confirm card's merge in `runConfirmTurn`. `tests/checkblock-one-row.test.js`.
66
+ - page rules: a label the page ALWAYS writes with a colon right after it ("Departure time: 10:24 AM.") is signed as the page writes it, whether the look gave "Departure time:" or "Departure time" (live run mwgsl7zx: after Revise the look dropped the colon, the time fact read no time and the person was asked to drop the rule). One owner, `withColon` in `src/pagerules.js`, a plain count of the page's text: every place that holds the label has a colon after it, else the word stays as given; the run's checker reads the same signed word, so drafting and run agree. Money and other value labels follow. `tests/time-label-colon.test.js` on the run's real page; the JFK-LAX map test and the golden port now expect the page's own word.
67
+ - **After an Agree, every view shows the replaced rule (live run mwgp0au6).** The chat message after the answers, the plan card's job lines and the saved `card.json` still read "~ not a red-eye" while the signed spec held "departure time between 05:00 and 23:59". The session's own lines now take the check list's rebuilt lines in place (one owner), `card` (what `card.json` stores) follows `state.card`, and the answers message is drawn after the rebuild.
68
+ - check list: one rule in, one rule out (hamr 2026-11-14, supersedes "an agreed answer is a new `~` line"; live run mwgjnrkz showed two `~` rules from one Agree). An Agree, or the person's own typed words, REPLACES the asked rule (same line, same rule number); the check list is the one owner and the job text, card box and signed rows are rebuilt from it (`jobLinesView`). Deleted: follower rows (`parent`, `matchFollow`, `followsOf`, `follow`, `pairWith`/`pairRule`, the "you agreed:" card tail), the F318 `sameRuleWords` dedupe in `answerQuestions`, and the picks record's `added` pairing. A counting rule whose own words are the check ("exactly 3 list items") is built as written with no question (`obviousCheck`), so a replaced rule never re-asks on a re-draft or Reuse. The ask text now says the answer "replaces that ~ rule". `!` and `$` rows already replace, never add beside (`onTheList`, carried choices are by row key). Test-proven; not seen live.
69
+ - rubric: a facts-only line that also asks for written words gets its note (live run mwgjlezb: "Write a one-sentence triage note ... quoting its title word for word" produced results and no note). At drafting, ONLY when a line is facts-only, ONE typed call (`src/noteask.js`, the destination call's provider/model/thinking-off, one retry, no repair) asks per such line whether it also wants written words; code decides: yes adds a `$` default row "line N writes a note — the note must be there" with a countable text part (brief FACTS AND TEXT, text-present floored), no changes nothing, a failed call shows a Not checked row "could not ask whether line N also wants a note". Test-proven; not seen live.
70
+ - rubric: the confirm card's "line N file ... must exist, be complete JSON, and carry text" lists "and carry text" only for a line whose signed close has a text-present check (`lineCarriesText`, derived from the close), not for every line.
71
+ - check list: a `!` phrase the AI adds (ledger-suggest) that is on the page but in none of its result rows ("Rome" in the search box and the "Track prices" toggles only) is Not checked with the look's own sentence ("... is on the page but in none of its results, so it is not checked."); it was built, offered under Suggested, signed, and the run went red every try. One owner for "what is a result row" (`resultQuotes`, `src/pagedry.js`) shared with the look's dry read; with no money fact on the line nothing is guessed. Found live by fwdloop (run walk-m16-flights run-1); the same class as the look's `phrase-no-result` (bareloop run mwg3pov1's flights checks).
72
+ - check list: a `~` money rule written in another currency than the page's fact ("under $150" on a euro price) is Not checked with the plain reason "the amount in your rule is in another currency than the page's prices"; it no longer falls back to the look's or the AI's reading and checks the number on the wrong currency. One spelling of the reason (`WHY_OTHER_CURRENCY`, `src/rubricstep.js`). Found live by fwdloop (run walk-m16-flights run-1, euro page); the same class as bareloop run mwg3pov1's money rows.
73
+ - panel: the Audit's checks fold (and the Result tab's checks box) show ONLY the run's latest close (the last `outer-close` or fix-loop `close-verdict`), not the `outer-close` precheck that runs before the fix loop; a run that went green in the fix loop no longer shows failed rows, and a stage the latest close never reached (it stops at its first red) reads grey "not reached" instead of the precheck's old result (live run mwg3pov1).
74
+ - panel: the Run summary "checks:" line ("N run · N passed · N failed") reads the same latest close as the Audit checks fold, through `checksOf`, not every step exit check plus the pre-fix-loop precheck; a stage the latest close never reached shows "N not reached" and counts as neither; the tries parenthesis is gone (live run mwg3pov1).
75
+
76
+ ### Added
77
+
78
+ - **Tickets + rubric goes through the check list (hamr 2026-11-14, piece 5; POC runs mwg70fd7 and mwg3pov1).** A Tickets job on the Rubric check has the ticket as its one input: the ticket line is the only input and is renumbered from 1 in ONE place (`cardInputs`, `src/panel/authorsession.js`), read by the fresh card, the session, the Destination check, the schedule and Reuse (`buildReuseSpec`) alike, so a Reuse no longer dies at signing with `spec-invalid (inputs[0].n expected 1)`. The drafting-time read takes the newest ticket the job has not run (`newestUnrunTicket`) and reads it (`readTicket`), with no ticket-number box; the frozen draft input is that ticket's text, and run start re-reads the same list row. Test seams: `createSession` deps `newestUnrunTicket` and `readTicketFn`.
79
+
80
+ ### Changed
81
+
82
+ - docs (hamr 2026-11-14, piece 7): `bareloop.context.md` (header, verdict classes, interview set, job block, rubric-on-files section, retired `judged-floor` note, Tickets + rubric section, exports list) and the README verification lines say what the code now does (no calibration, no judge decides, no fit check, no PASS/FAIL examples); PRD v1.134 contract lines ticked.
83
+ - **An old job (PASS/FAIL examples or the old judge) is not reused or bundled (hamr 2026-11-14, piece 6).** A signed spec with a `~ PASS:` / `~ FAIL:` job-line rule or a `judged-floor` stage refuses in plain words ("This job was made with PASS/FAIL examples or the old judge, which are gone. Draft it fresh.") at Reuse (the panel route and the session), at a scheduled fire (it is a Reuse), and at `bareloop run <bundle>`, before any key, repo or copy is touched. One reader: `oldJobWords` (`src/job.js`). Old runs are not repaired.
84
+ - **Rubric on a repo is refused at $0 (hamr 2026-11-14, piece 4).** A Deterministic check is anything a machine can check, so a repo job is Deterministic; Rubric on a repo is allowed only for a Tickets job. A rubric card whose Inputs line 1 is a repo, and a `soft-green` pick at the CLI interview (it always takes a repo), are refused before anything is made or asked: "A repo job uses a Deterministic check."
85
+ - panel: the Audit checks fold header is lowercase like the other part rows (hamr 2026-11-13); README walkthrough GIF regenerated for the check block.
86
+ - Dependency: bareguard ^0.36.0 (check helpers, ruling #82; existing rubric fingerprints and verdicts unchanged — peer-tested on 90 saved plans and 35 verdicts).
87
+
8
88
  ## [0.56.0] - 2026-11-13
9
89
 
10
90
  ### Added
package/README.md CHANGED
@@ -25,7 +25,7 @@ bareloop finds that path for you. Write your problem in plain language, as steps
25
25
  The harness (budget, time, what it may touch) and the judge both sit **outside** the agent loop. The agent can't judge itself. Every run is recorded, so you can audit it and replay it at $0 and see exactly what happened.
26
26
 
27
27
  <p align="center">
28
- <img src="https://raw.githubusercontent.com/hamr0/bareloop/main/docs/media/panel-walkthrough.gif" alt="The bareloop panel: write a job, the runs list, a rubric run waiting for Accept, its Result, Audit and signed Job, then a repo run ready to merge" width="900">
28
+ <img src="https://raw.githubusercontent.com/hamr0/bareloop/main/docs/media/panel-walkthrough.gif" alt="The bareloop panel: write a job, the runs list, a rubric run with a fixed try, its Result, the Audit checks fold and the signed Job with its check rows, then a repo run ready to merge" width="900">
29
29
  </p>
30
30
 
31
31
  ## The machine that makes machines
@@ -36,7 +36,7 @@ Like a CNC machine, bareloop builds the thing that does the work, then lets a ga
36
36
 
37
37
  **Guide rails, not just guardrails.** Guardrails stop the agent going off the road: a money cap, a time cap, and a read/write fence it can't move. Guide rails keep it on the road: each new worker hears what earlier tries read, searched and changed, diffs included; two misses redraw the plan; past lessons go to the planner; an unclear job gets up to 3 questions, not a guess.
38
38
 
39
- **Verification.** The check sits outside the agent's loop: code first, a judge only where code can't decide, and the judge only points while code makes the call. A red goes back with the real error to be fixed, until it's green or out of budget.
39
+ **Verification.** The check sits outside the agent's loop: code decides. Anything code can't check is listed as Not checked, and you judge it at Accept. A red goes back with the real error to be fixed, until it's green or out of budget.
40
40
 
41
41
  **Evidence.** Every call, every allowed or blocked read and write, every check result and every cent is recorded. You can audit it in the panel and replay it at $0. Research facts are quoted from saved sources.
42
42
 
@@ -45,7 +45,7 @@ Like a CNC machine, bareloop builds the thing that does the work, then lets a ga
45
45
  **Deterministic (code)**: code decides. Tests pass, the linter is quiet, the build exits clean. If no check exists yet, say what you need and the agent builds that check first.
46
46
  > *"Fix the lint errors."* · *"Add doc comments to every function."* · *"Split this 3,000-line file without breaking a test."*
47
47
 
48
- **Rubric (research)**: it searches, collects, compares and hands you a report where every fact is quoted from a saved copy of its source. A judge only points at the quotes, and fixed rules decide.
48
+ **Rubric (research)**: it searches, collects, compares and hands you a report where every fact is quoted from a saved copy of its source. An AI only points at the page's own words, and fixed rules in code decide.
49
49
  > *"Find flights LAX→SFO under $400, nonstop, no red-eye."* · *"Compare these 5 laptops on price, weight and warranty."*
50
50
 
51
51
  ## What it won't do