bareloop 0.24.0 → 0.25.0

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.
package/CHANGELOG.md CHANGED
@@ -5,7 +5,103 @@ 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
- ## [Unreleased]
8
+ ## [0.25.0] — 2026-09-12
9
+
10
+ ### Added
11
+
12
+ - **The source front door** (PRD item 33/M2) — a plain folder, a plain file, or one
13
+ `http(s)` URL now runs without a `--patient` git checkout: `prepareSource`
14
+ (`src/source.js`) freezes it into a hidden, scratch git tree bareloop owns (the person
15
+ never sees the repo), `proveDestination` proves a drop-off file is landable at $0 before
16
+ any token spends, and `copyOut` delivers the run's own output there once, on a minted
17
+ green, never overwriting. New CLI `scripts/prep-source.mjs`. `scripts/run-u.mjs` wires
18
+ both: a manifest-declared destination refuses before any token, and copies out on a
19
+ green (a refused copy-out never changes the verdict). The destination is a PER-RUN value
20
+ recorded in the manifest (`<into>/source.json`), never a signed job-spec field — a job
21
+ spec is a repeatable shape, signed once; nothing was added to `job.js`/`jobSpecHash`.
22
+ Exported: `prepareSource`, `proveDestination`, `copyOut`, `readSourceManifest`,
23
+ `frontDoorFromManifest`. `git()` (`src/kinds.js`) is now exported for reuse by this
24
+ module rather than re-spawned a second way.
25
+
26
+ ### Fixed
27
+
28
+ - **Source front door (PRD item 33/M2 review):** `output` is now validated the ONE way
29
+ everywhere it is read or written (new `output-invalid` refusal) — a bare
30
+ `resolve(tree, output)` previously let a shape like `output/../../x` escape the frozen
31
+ tree; a hand-edited manifest carrying such a value is now treated as no front door at
32
+ all by `frontDoorFromManifest`, never handed to `copyOut` unvalidated. `prepareSource`
33
+ now proves the destination (`proveDestination`) BEFORE creating `into`, so a refused
34
+ destination leaves nothing on disk. `copyOut` takes an optional `into` (the run's
35
+ scratch root); `scripts/run-u.mjs`'s two call sites now pass `dirname(wd)` instead of
36
+ `wd`, so a destination inside the scratch area but outside the frozen tree is caught as
37
+ `destination-contained` instead of slipping through.
38
+ - **Source front door hard-line defect (PRD item 33/M2, live-proven):** every frozen
39
+ file's content — a folder, a single file, or a fetched URL body — is now scanned for a
40
+ known secret shape (the same `scanSecrets` inventory, `src/validate.js`) BEFORE
41
+ anything is written under `into`; a hit refuses `source-carries-secret`, naming the
42
+ file(s) and pattern name(s) only, never the matched text, and leaves nothing on disk.
43
+ Previously the door scanned only the URL string, so a plain folder or file carrying a
44
+ real API key (e.g. an `.env`) was frozen into the tree and committed to the hidden git
45
+ seed untouched.
46
+ - **Source front door, M2b review fixes 2–8 (PRD item 33):** `.env`/`.env.*` refused by
47
+ NAME whatever it contains (`source-env-file`), belt-and-braces beside the content scan;
48
+ a per-file `MAX_BUFFER` ceiling for folder and single-file sources (`source-file-
49
+ oversize`; repo sources exempt); the delivered file now carries the date
50
+ (`profile-2026-09-12.md`, `-2`/`-3` the same day, cap 99) — new exports
51
+ `datedDestination`/`pickDelivery`, used by both `proveDestination` and `copyOut` so the
52
+ two can never disagree; `copyOut` returns `{bytes, sha256, path}` and `run-u.mjs`
53
+ spine-records the real landed path (`destination-written {path, declared, bytes,
54
+ sha256}`); a URL's redirects are followed but never invisible — the final URL rides in
55
+ the manifest (`finalUrl`) and is printed by `prep-source`, and is secret-scanned like
56
+ the typed one; a nested `.git` anywhere below the root refuses `source-nested-repo`,
57
+ `git add` runs `-f` so a source `.gitignore` cannot drop a file from the seed, and the
58
+ seed is read back with `git ls-tree` and diffed against the manifest's file list
59
+ (`source-seed-incomplete` on a shortfall); a git repo root is now an allowed source
60
+ (`kind: 'repo'`) — copied with its history, working files at the tree root, seed on top
61
+ of existing history; a `.git` FILE (linked worktree/submodule) refuses
62
+ `source-is-linked-worktree`; `prep-source` prints a folder-blast-radius note (mirrored
63
+ in `bareloop.context.md`).
64
+ - **Repo-source hard-line defect (live-proven during the M2b 7 review):** a copied
65
+ `.git/hooks` carried the SOURCE repo's own `pre-commit`/`commit-msg`/`post-commit`
66
+ scripts, and `git commit` ran them — arbitrary code from the source repo, executed
67
+ inside bareloop's own process, before a single token spent. Fixed: the copied hooks
68
+ directory is stripped after the `.git` copy, and every git call the door makes pins
69
+ `core.hooksPath` to a path that is never created, so a `core.hooksPath` set in the
70
+ copied `.git/config` cannot reopen the hole either.
71
+
72
+ - **Source front door D1/D2/D3 rework (PRD item 33, hamr's rulings, 2026-09-12):**
73
+ - **D1 — repo sources copy only what git tracks** (hamr, verbatim: "copy only what git
74
+ tracks"; closes F164/F165): repo sources are enumerated with `git ls-files --stage` in
75
+ the source, never a filesystem walk; each tracked path's working-tree content is
76
+ frozen, and nothing untracked or gitignored is ever a candidate (npm
77
+ `node_modules/.bin/*` symlinks, this repo's own `.claude/`, a gitignored `.env`,
78
+ none of it reaches the tree or the seed). `git add -A` drops `-f` for repo sources
79
+ (kept for plain folder/file/URL sources). A gitlink (a real submodule) refuses
80
+ `source-nested-repo`; a tracked symlink refuses `source-symlink` only when it
81
+ resolves outside the source root — one staying inside is copied verbatim as a link.
82
+ - **D2 — secrets scanned inside binary content too** (closes the F166 residual): every
83
+ frozen file's content is scanned for secrets, binary files included, decoded `latin1`
84
+ (byte-preserving) instead of `utf8` — the same `SECRET_PATTERNS` inventory, no second
85
+ pattern list.
86
+ - **D3 — destination is a DIRECTORY, never a filename** (hamr's ruling, 2026-09-12): it
87
+ may already exist, need not be empty, and may sit inside the source itself
88
+ (`destination-in-source` is gone). A job may produce more than one file; `copyOut`
89
+ delivers every non-empty file under `output/` (excluding `output/.gitkeep`), each
90
+ under its own dated name, and returns `{files: [{path, bytes, sha256}]}` instead of a
91
+ single object. The single-file `output` manifest field and `output-invalid` validator
92
+ are removed — the escape vector they guarded is now closed structurally, since
93
+ `copyOut` reads `tree/output/` itself rather than a manifest-declared path. A repo
94
+ source's `destination` is recorded as the declared write fence and never proven or
95
+ copied into (`frontDoorFromManifest` returns `null` for a repo manifest) — wiring it
96
+ into the signed `writeScope` field is left to a later milestone. New refusal codes:
97
+ `destination-not-directory`, `destination-not-writable`, `destination-invalid`
98
+ (repo-destination type guard); retired: `output-invalid`,
99
+ `destination-output-required`, `destination-in-source`. `scripts/prep-source.mjs`
100
+ drops `--output`; `scripts/run-u.mjs`'s copy-out call site emits one
101
+ `destination-written` per delivered file.
102
+ - **D4 — measured a real 15 MB repo** (`~/PycharmProjects/adaptlearn`, the repo that
103
+ broke F164): 1296 files, 3.16 MB copied, 0.89s elapsed, ~84 MB maxRSS, no gitignored
104
+ or untracked content reached the tree or the seed. No size cap added.
9
105
 
10
106
  ## [0.24.0] — 2026-09-10
11
107
 
@@ -2907,6 +2907,183 @@ provider but `anthropic-api`, spine bundling for `formatReplay` on the minting r
2907
2907
  `runid` is noted), and tarball packing — a bundle is a plain directory; use `npm pack` on it
2908
2908
  yourself if you want a tarball.
2909
2909
 
2910
+ ### The source front door — a plain folder/file/URL, no `--patient` (PRD item 33/M2, `src/source.js`)
2911
+
2912
+ A repo job still gives `runJob` a `--patient` checkout it owns. A NON-code job (a resume, a
2913
+ CSV, a page) has no repo to point at — this door is what lets it run anyway, WITHOUT the
2914
+ person ever seeing a git repo (H1's fix): the source is frozen into a hidden, scratch git
2915
+ tree bareloop creates and owns, before any token spends, so the run and its close judge the
2916
+ SAME bytes.
2917
+
2918
+ ```js
2919
+ import { prepareSource, proveDestination, copyOut, readSourceManifest, frontDoorFromManifest } from 'bareloop';
2920
+
2921
+ const prepared = await prepareSource({
2922
+ source: '/home/me/resume.md', // an http(s) URL, an existing file, or an existing plain folder/repo
2923
+ into: '/home/me/.bareloop-jobs/j1', // must NOT already exist — fresh, never reused
2924
+ destination: '/home/me/profile-drafts', // optional — a DIRECTORY (D3), need not exist or be empty
2925
+ });
2926
+ // prepared.tree === '/home/me/.bareloop-jobs/j1/tree' — hand THIS to runJob as `workdir`
2927
+ // prepared.manifest.seed — the commit `seedAtHead` will read at run start
2928
+ ```
2929
+
2930
+ Or the CLI: `node scripts/prep-source.mjs --source <path-or-url> --into <dir> [--destination
2931
+ <absolute-directory>]` — prints the tree path, the seed, and the exact next command.
2932
+
2933
+ > **If your source is a folder, read this first (hamr's ruling):** make a new folder, put
2934
+ > only the file(s) this job needs in it, point bareloop at that — never your original
2935
+ > folder. Everything in the folder you name is frozen, read, and paid for. `prep-source`
2936
+ > prints the same line on every folder prep, and the panel (N6) shows it beside the Source
2937
+ > field.
2938
+
2939
+ **Layout.** `<into>/tree/input/…` is the frozen copy; `<into>/tree/output/` is empty (a
2940
+ `.gitkeep` seeds it) and is where the run writes; `<into>/source.json` is the manifest,
2941
+ OUTSIDE the tree — no worker verb can ever read or edit it. One hidden commit, authored
2942
+ `bareloop <bareloop@localhost>` with `commit.gpgsign` pinned off — never the operator's
2943
+ identity, never global git config (this repo's own tests run with an empty `HOME`).
2944
+
2945
+ **What is admitted today (hamr: "start with already supported files").** A URL is fetched
2946
+ ONCE with plain `fetch` — no credentials, no cookies, no custom headers — under
2947
+ `PROVIDER_TIMEOUT_MS` and capped at `MAX_BUFFER`, both reused from elsewhere in this library,
2948
+ never a second number. Content-type must be `text/*`, `application/json`, or
2949
+ `application/xml`. A local file or folder must be plain text (a NUL byte in the first 8KB of
2950
+ any file refuses the whole prep, naming every offending file) and must contain no symlinks
2951
+ (named, never followed). Each file has its OWN `MAX_BUFFER` ceiling (`source-file-oversize`)
2952
+ — there is no folder TOTAL cap, hamr's call. A file named `.env` or `.env.*` is refused by
2953
+ NAME whatever it contains (`source-env-file`), belt-and-braces beside the content scan below.
2954
+ A `.git` found anywhere BELOW the root is a nested repo and refuses (`source-nested-repo`):
2955
+ its objects are a second history nobody declared, and the seed could not hold them honestly.
2956
+ PDF and Word input have no reader here (hole H7, PRD item 33) — logged, not built.
2957
+
2958
+ **A git repo IS an allowed source** (hamr: yes — the PR-review job). It is COPIED with its
2959
+ history, never used in place (the patients-are-copies rule): the working files land at the
2960
+ TREE ROOT rather than under `input/`, `.git` is copied verbatim, `output/` is added, and the
2961
+ seed commit goes on TOP of the existing history — so `manifest.kind === 'repo'` and the
2962
+ worker gets a real repo to review. Its guards (the input stays untouched) arrive with M4. A
2963
+ `.git` that is a FILE, not a directory — a linked worktree or a submodule — refuses
2964
+ `source-is-linked-worktree`: its real git directory lives elsewhere, and a copy would
2965
+ silently commit into the original. Text-only and the per-file ceiling do not apply to a repo
2966
+ source (a real repo legitimately carries binaries).
2967
+
2968
+ **A repo source copies only what git tracks** (D1 rework, hamr's ruling verbatim: "copy only
2969
+ what git tracks" — closes F164/F165). Files are enumerated with `git ls-files --stage` in the
2970
+ SOURCE, never a filesystem walk: each tracked path's WORKING-TREE content is frozen (so
2971
+ uncommitted edits to a tracked file come along), and nothing untracked or gitignored is ever a
2972
+ candidate — real npm `node_modules/.bin/*` symlinks, this repo's own `.claude/`, a gitignored
2973
+ `.env`, none of it is read or written. `git add -A` drops `-f` for a repo source (kept for a
2974
+ plain folder/file/URL, where a `.gitignore` has no honest meaning yet). A gitlink entry (a real
2975
+ submodule) refuses `source-nested-repo`; a tracked symlink refuses `source-symlink` only when
2976
+ it resolves OUTSIDE the source root — one that stays inside is legal tracked content, copied
2977
+ verbatim as a link, never dereferenced. Measured live against a real 15 MB repo carrying
2978
+ `node_modules` (`~/PycharmProjects/adaptlearn`, the repo that broke F164): 1296 files, 3.16 MB,
2979
+ 0.89s elapsed, ~84 MB maxRSS — no gitignored or untracked content reached the tree or the seed.
2980
+
2981
+ **Every frozen file's content is secret-scanned, binary files included** (D2, closes the F166
2982
+ residual). A binary file (a NUL byte in its first 8KB) used to be hashed but never scanned —
2983
+ now it runs through the SAME `SECRET_PATTERNS` inventory, decoded `latin1` (byte-preserving)
2984
+ instead of `utf8`, so an ASCII key embedded in a compiled artifact is still found. This only
2985
+ matters where binaries are admitted at all (a repo source today; a plain folder/file still
2986
+ refuses binary content outright via `source-not-text`, unchanged).
2987
+
2988
+ **A URL's redirects are followed but never invisible.** The URL the body actually came from
2989
+ rides into the manifest as `finalUrl` and is printed by `prep-source` whenever it differs
2990
+ from what was typed, so a link that quietly lands on a login or error page is something a
2991
+ person can see before a token spends. The final URL is secret-scanned the same way the typed
2992
+ one is.
2993
+
2994
+ **The seed holds every frozen file, and that is CHECKED, not hoped.** `git add` runs with
2995
+ `-f`, so a `.gitignore` inside the source cannot drop files from the seed (a file present on
2996
+ disk but absent from the seed would read to a close as the worker having written it); then
2997
+ the seed is read back with `git ls-tree -r --name-only HEAD` and diffed against the manifest's
2998
+ own file list by name. A shortfall refuses `source-seed-incomplete`.
2999
+
3000
+ **Every frozen file's content is scanned for a known secret shape before anything is
3001
+ written under `into`** (a folder, a single file, or a fetched URL body — the SAME inventory
3002
+ `scanSecrets` reads, `src/validate.js`; live-proven fix — the door used to scan only the URL
3003
+ string, so a plain `.env` with a real API key was frozen into the tree and committed to the
3004
+ hidden git seed untouched). A hit refuses `source-carries-secret`, naming the file path(s)
3005
+ and the pattern name(s) only — the matched text itself never appears in the refusal, the
3006
+ manifest, or any file, and nothing is written to disk (`into` does not exist afterward). This
3007
+ is a shape-based check, not a filename denylist: a secret whose shape is not in the inventory
3008
+ (a plain database password in a `.env`, say) still passes through — a named, accepted
3009
+ residual, not this fix's job to close.
3010
+
3011
+ **Every refusal is a named `{stop, code}`, never a throw and never silent:** `source-
3012
+ unreadable`, `source-symlink`, `source-env-file`, `source-file-oversize`,
3013
+ `source-nested-repo`, `source-is-linked-worktree`, `source-seed-incomplete`, `source-git-failed`, `source-not-text`, `source-fetch-failed`,
3014
+ `source-fetch-timeout`, `source-fetch-oversize`, `source-carries-secret`, `into-exists`,
3015
+ `destination-invalid`, `destination-not-absolute`, `destination-not-directory`,
3016
+ `destination-not-writable`, `destination-contained`, `destination-parent-missing`,
3017
+ `destination-parent-unwritable`, `destination-exists`, `destination-output-missing`,
3018
+ `destination-write-failed`, `source-manifest-unreadable`, `source-manifest-invalid`. A
3019
+ refusal never throws, never overwrites a person's file, and never lands a destination inside
3020
+ the run's own scratch tree.
3021
+
3022
+ **Destination is a DIRECTORY, never a filename** (D3 rework, hamr's ruling, 2026-09-12,
3023
+ condensed in `docs/product/ITEM33-BUILD.md`). It may already exist and need not be empty; it
3024
+ may sit inside the source itself (the same-dir case: "i can limit your actions to a certain
3025
+ fix in a certain dir and your output will also be there" — `destination-in-source` is gone). A
3026
+ job may produce MORE THAN ONE file (his example: a flight search producing one sheet for
3027
+ "SFO→LAX red-eye" and another for "SFO under $700"); the agent names them, and there is no
3028
+ `output` field in the manifest any more — `copyOut` always reads `tree/output/` itself, never
3029
+ a manifest-declared path, which closes the old escape vector (`output/../../x`) structurally
3030
+ rather than by validation.
3031
+
3032
+ **Destination is proven at prepare time, not just in-run** (hamr's ruling: source AND
3033
+ destination are proven at job start, $0): when `destination` is given, `prepareSource` calls
3034
+ `proveDestination` BEFORE creating `into` — a refused destination leaves nothing on disk.
3035
+ `proveDestination` only proves the directory is USABLE (exists as a directory and is writable,
3036
+ or its parent exists and is writable so it can be created) — it cannot foresee individual
3037
+ delivered FILE names, since those are not known until the run actually produces them.
3038
+ `copyOut` takes an optional `into` (the run's scratch root, `<into>`, of which `tree` is the
3039
+ `tree/` subdirectory; defaults to `tree` for direct callers with no `into`) so a destination
3040
+ sitting inside the scratch area but outside the frozen tree — e.g. `<into>/profile.md` — is
3041
+ also caught as `destination-contained`, not just one strictly inside `tree`. `run-u.mjs`'s two
3042
+ call sites pass `dirname(wd)` (`wd` IS `<into>/tree`) as `into` for exactly this reason.
3043
+
3044
+ **A `kind: 'repo'` source's destination is a DIFFERENT thing** (hamr: "destination if for code
3045
+ on a green then that would be the place agent allowed to do changes") — it names the WRITE
3046
+ FENCE inside the copied repo, the signed `writeScope` field's job (`src/job.js:363`), never a
3047
+ filesystem drop-off point. It is recorded as declared, never proven by `proveDestination` and
3048
+ never handed to `copyOut` (`frontDoorFromManifest` returns `null` for a repo manifest).
3049
+ Wiring it into `writeScope` is M3/M4's job, not this door's.
3050
+
3051
+ **Every non-empty file under `output/` is delivered, each under its OWN dated name, and
3052
+ nothing is ever overwritten.** A file named `profile.md` LANDS as `profile-2026-09-12.md`
3053
+ inside the destination directory; a second delivery of the same name the same day lands as
3054
+ `profile-2026-09-12-2.md`, then `-3` (the work-branch collision rule), up to `-99` before
3055
+ `destination-exists`. An empty file is silently skipped (not a delivered result); the batch
3056
+ refuses `destination-output-missing` only when NOTHING non-empty exists under `output/` at
3057
+ all (excluding the `output/.gitkeep` seed placeholder). `datedDestination(destination, now,
3058
+ n)` and `pickDelivery(destination, now?)` are exported and unchanged from M2b — `copyOut`
3059
+ calls them per file now, at delivery time, rather than once against a single declared path.
3060
+ `copyOut` returns `{files: [{path, bytes, sha256}]}` and `run-u.mjs` spine-records ONE
3061
+ `destination-written` per delivered file — the declared directory is not the path anything
3062
+ wrote.
3063
+
3064
+ **The destination is a PER-RUN value, never a signed job-spec field.** A job spec is a
3065
+ repeatable SHAPE (goal, checks, judge rules), signed once; source and destination vary per
3066
+ run — the same way `bareloop run <bundle> --repo <path>` varies the repo instance without
3067
+ touching the bundle's own signature. So `job.js`/`jobSpecHash` carry nothing about it: the
3068
+ run's own instance lives entirely in the manifest `prepareSource` writes.
3069
+
3070
+ `scripts/run-u.mjs` reads that manifest via `readSourceManifest(dirname(workdir))` (`workdir`
3071
+ IS `<into>/tree`, so its parent is `<into>`) and reduces it with `frontDoorFromManifest`. No
3072
+ manifest → no front-door behaviour at all — every existing repo-patient job is untouched. A
3073
+ manifest that exists but cannot be read or parsed is a named stop, never a silent skip. When
3074
+ a manifest declares a destination, `proveDestination` runs BEFORE any token spends (a job
3075
+ that cannot land its output should never buy a worker turn first), and `copyOut` runs once,
3076
+ on a minted `green` — the ONE outcome string a graded close ever mints; `soft-green` jobs
3077
+ mint the same string, since the class lives on `spec.verdictType`, never on `runJob`'s
3078
+ return value. A refused copy-out is spine-recorded (`destination-refused {code, detail}`)
3079
+ beside the green it could not deliver and NEVER changes that verdict — a delivery failure is
3080
+ not a grading failure.
3081
+
3082
+ Not wired yet, named rather than silently absent: `bareloop run` (the exported bundle CLI,
3083
+ `src/cli.js`) still requires `--repo`; a bundled front-door job is parked past PRD item
3084
+ 33/M5. The intake form that lets a person fill in Source/Destination through the authoring
3085
+ interview (rather than the CLI above) is item 33/M3, not yet built.
3086
+
2910
3087
  ## Architecture
2911
3088
 
2912
3089
  Three layers. An **outer shell** (dumb, permanent): per-run budget cap via bareguard,
package/package.json CHANGED
@@ -1,6 +1,6 @@
1
1
  {
2
2
  "name": "bareloop",
3
- "version": "0.24.0",
3
+ "version": "0.25.0",
4
4
  "description": "Workflows that earn their own design, with receipts — agent-authored workflows for repeated, long, verifiable jobs that improve across runs under an un-gameable gate.",
5
5
  "type": "module",
6
6
  "main": "./src/index.js",
package/src/index.js CHANGED
@@ -285,3 +285,12 @@ export {
285
285
  // exists so an adopter driving `bareloop run` from their own script can do so
286
286
  // without spawning `bin/bareloop.mjs` as a child process.
287
287
  export { main as cliMain } from './cli.js';
288
+ // PRD item 33/M2 (`docs/product/ITEM33-BUILD.md`) — the source front door. A
289
+ // plain folder, a plain file, or one URL frozen into a hidden git tree before
290
+ // any token spends (H1's fix: the person never sees the repo), and the
291
+ // destination proven/copied at the two moments a runner needs them ($0
292
+ // preflight, and once on a minted green). Exported because `scripts/prep-
293
+ // source.mjs` and `scripts/run-u.mjs` are two separate call sites reaching
294
+ // for the SAME functions — an adopter building a third has to reach the same
295
+ // module, not a re-derivation of it.
296
+ export { prepareSource, proveDestination, copyOut, readSourceManifest, frontDoorFromManifest } from './source.js';
package/src/kinds.js CHANGED
@@ -485,9 +485,15 @@ async function sh(cmd, args, { cwd, env: declared = {}, timeoutMs = DEFAULT_TIME
485
485
  /**
486
486
  * git, read-only, in a repository. Faults come back as `ok:false` with the
487
487
  * stderr — a caller decides whether that is a stop.
488
+ *
489
+ * Exported (PRD item 33/M2, `src/source.js`) so the hidden-git seed a plain
490
+ * folder/file/URL job runs against goes through the SAME spawn primitive as
491
+ * every other git read in this file — a second `spawn('git', …)` for the
492
+ * source front door would be a second instrument with its own timeout and
493
+ * buffer ceiling to keep in sync with this one.
488
494
  * @param {string} cwd @param {string[]} args
489
495
  */
490
- async function git(cwd, args) {
496
+ export async function git(cwd, args) {
491
497
  const r = await sh('git', args, { cwd, timeoutMs: 120_000 });
492
498
  // git refusing IS "the close cannot run": the changed-set primitive is not an
493
499
  // optional enrichment, so a non-zero git carries FAILED rather than inheriting