@vaultcompass/conductor 0.3.0 → 0.4.5

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/README.md CHANGED
@@ -56,6 +56,14 @@ become the annoying part:
56
56
  npm install -g @vaultcompass/conductor
57
57
  ```
58
58
 
59
+ **Globally, and that is the guidance for CI too.** A devDependency is fine
60
+ beside it for your own pre-commit hook, and nothing about a local run changed.
61
+ It is not what gates your pull requests: on a pull-request run the umbrella
62
+ never looks in `node_modules/.bin` at all, because what is installed there is
63
+ chosen by the head's own manifest and lockfile. The Action installs all four
64
+ packages itself, globally, at versions pinned in your workflow file. See "The
65
+ pull-request trust boundary" and "The Action" below.
66
+
59
67
  From the repository root, look before you write:
60
68
 
61
69
  ```
@@ -304,6 +312,44 @@ where a head-committed `node_modules/.bin/<gate>` shadows the real gate,
304
312
  because resolution prefers the repository's own copy over PATH so that a
305
313
  project pin beats a global install.
306
314
 
315
+ **On a pull-request run `node_modules/.bin` is not consulted at all.** That
316
+ second shape is closed one step earlier than the first, and this is why: what
317
+ is installed there is chosen by the head's own manifest and lockfile, so it is
318
+ never approved by any ref, and a repository whose gates are devDependencies
319
+ would otherwise meet the program refusal on every ordinary pull request. So
320
+ that location is simply not searched when a trust base is set. PATH and an
321
+ absolute `command:` remain, and both still go through the program check below.
322
+ The skip is never silent: one line in the full report and one
323
+ `conductor/node-modules-skipped` note in the SARIF log name each gate and the
324
+ copy that was not taken, and a gate with nothing on PATH either is
325
+ could-not-run with `npm install -g` and the Action's version input named as
326
+ the remedy. **Outside pull-request mode nothing changes**: your own checkout
327
+ and your pre-commit hook still run the pin, which is what `pnpm exec` does in
328
+ the same repository.
329
+
330
+ **The versions that judge a pull request are pinned in the workflow file**,
331
+ and that is the point: the pull request can change its own lockfile,
332
+ manifest and `node_modules`, and none of those decide which conductor and
333
+ which gates run over it any more. Be clear about what that does not cover.
334
+ On a `pull_request` event GitHub runs the workflow as it is in the pull
335
+ request's merge commit, so a pull request that edits the workflow file can
336
+ change the pins or remove the gates step, exactly as it could rewrite any
337
+ other step in your CI. The boundary here is against the tree choosing its
338
+ own judge; the control for a workflow edit is branch protection on your
339
+ base branch with review required for `.github/workflows`, and no action can
340
+ provide that for you. Only `pull_request_target` runs the base branch's copy
341
+ of a workflow, and that event exposes the base's secrets to the pull
342
+ request's code, which is the wrong trade for a gate over untrusted changes.
343
+
344
+ One part of that is closed, and only one. Because the pins are written in a
345
+ file the pull request controls, a pull request could otherwise pin a gate
346
+ **backward** to a published version that predates the rule which would have
347
+ caught it: an exact version, so the validate step's shape check accepts it,
348
+ and a change that reads as ordinary version management. **On a pull request
349
+ the action now refuses any of the four `*-version` inputs naming a version
350
+ below the one the action tag ships.** Pinning forward is still accepted.
351
+ See "The Action" below.
352
+
307
353
  So on a pull-request run a gate's program must be **outside the working tree**
308
354
  (on PATH, or an absolute `command:` elsewhere on the machine), or else meet
309
355
  **both** of these:
@@ -341,12 +387,13 @@ the exit code even where the policy sets `enforce: false`, because the gate
341
387
  produced no findings to un-enforce: the umbrella declined to run a program
342
388
  the pull request chose.
343
389
 
344
- Nothing under `node_modules` is ever base-approved, and that is the right
345
- answer rather than a limitation: what is there is chosen by the head's own
346
- manifest and lockfile and installed by a step that runs before the gates, so
347
- git has no record of those bytes at either ref. A pull request that edits its
348
- lockfile to pull a different build of a gate has chosen its own judge just as
349
- surely as one that commits a stub. **Vendoring a gate still works**, on those
390
+ Nothing under `node_modules` is ever base-approved, which is why the rule
391
+ above does not look there at all. Were it to look, the answer would always be
392
+ the same refusal: what is there is chosen by the head's own manifest and
393
+ lockfile and installed by a step that runs before the gates, so git has no
394
+ record of those bytes at either ref. A pull request that edits its lockfile to
395
+ pull a different build of a gate has chosen its own judge just as surely as
396
+ one that commits a stub. **Vendoring a gate still works**, on those
350
397
  terms: put it in its own directory, keep everything it needs inside that
351
398
  directory, and leave the whole directory alone in a pull request the gate is
352
399
  meant to judge. Changing it is not forbidden, it just has to land on the base
@@ -489,14 +536,75 @@ no severity threshold of its own here either.
489
536
  `action.yml` at the root is a composite action that runs the gates at the
490
537
  `ci` stage and writes one SARIF log for the caller to upload.
491
538
 
492
- It **installs nothing**, on purpose. It runs the conductor your repository
493
- already depends on, so the version gating your pull requests is the one your
494
- lockfile pins rather than whatever the registry serves that morning; instead
495
- of installing, the action checks for `node_modules/.bin/conductor` and fails
496
- with a sentence saying to add it as a devDependency. `--base` is passed only when the
497
- `base-ref` input names one; left empty, the umbrella reads `GITHUB_BASE_REF`
498
- itself and treats an empty value as "not a pull request", which is what a
499
- push build wants.
539
+ It **installs all four packages itself**, and outside the workspace. Until
540
+ 0.4.0 it installed nothing and ran the `node_modules/.bin/conductor` your
541
+ repository already had, which put every program in the family inside the tree
542
+ under judgment: the install step runs the head's lockfile, so a pull request
543
+ could repoint any gate, or the umbrella itself, at a build it controls while
544
+ leaving the version number alone.
545
+
546
+ Now four inputs name exact versions, `conductor-version`,
547
+ `dep-guard-version`, `vault-guard-version` and `intent-guard-version`, each
548
+ defaulting to the version this release pins. A validate step refuses anything
549
+ that is not an exact version, a range and `latest` included: the version that
550
+ judges a pull request has to be a decision taken on the base branch rather
551
+ than one the registry takes on the morning of the run. An install step then
552
+ runs `npm install -g --ignore-scripts` under the runner temp, never into the
553
+ workspace, and prepends that bin directory to `PATH`, so the run step invokes
554
+ `conductor` off `PATH` and each gate resolves by name the same way. The install
555
+ is **unconditional**, on push and `pull_request` alike, so there is one code
556
+ path rather than one that matters and one that nobody exercises.
557
+
558
+ **On a pull request those four inputs may not pin backward.** The shape check
559
+ asks whether an input is an exact version; it says nothing about which one, and
560
+ on a `pull_request` event the workflow file comes from the head, so the pins are
561
+ written by the pull request being judged. Once a gate has two published versions
562
+ that is a bypass with an innocent shape: pin back to the release that predates
563
+ the rule which would have caught the change, and the change picks the rules it
564
+ is judged by. So where `GITHUB_BASE_REF` is set, the validate step refuses an
565
+ input naming a version below the one the action tag ships, naming both numbers,
566
+ and the fix is to **remove the input**: the default is that version. Pinning
567
+ **forward** is still accepted, on an assumption the rule does not enforce, that
568
+ a newer gate is at least as strict; nothing bounds how far forward you pin.
569
+
570
+ Three things follow from that, stated because the summary is wider than the rule.
571
+ It fires on fork pull requests too, where the base repository's workflow file is
572
+ the one that runs, so a deliberate backward pin you wrote yourself will refuse
573
+ every fork run until you remove it. Push runs are out of scope rather than safe:
574
+ a push to an unprotected branch runs that branch's own workflow file, written by
575
+ the same author. And `merge_group` runs are not covered at all: `GITHUB_BASE_REF`
576
+ is set on `pull_request` and `pull_request_target` only, so it is empty in a
577
+ merge queue even though the queue branch carries the pull request's commits and
578
+ its workflow file. If your only required check runs on `merge_group`, this rule
579
+ does nothing for you; keep the `pull_request` run required as well and the pin is
580
+ caught before the queue. Neither this nor anything else in `action.yml`
581
+ replaces branch protection with review required for `.github/workflows`, which
582
+ is still the only control over a pull request that edits the workflow.
583
+
584
+ `--ignore-scripts` is there because the step holds the job's token and the four
585
+ things it installs decide whether a pull request may merge; without it every
586
+ package in the resolved tree would run code on the runner. The step then runs
587
+ `npm audit signatures` over what it installed.
588
+
589
+ **What that verification proves, and what it does not.** It asks the registry
590
+ for each name and version in the tree, the four gates included, and checks the
591
+ signature served back, so an unpublished, replaced or unsigned package fails the
592
+ step. It does **not** read the installed files, so it will not detect a tampered
593
+ install; it does **not** defeat a compromised registry, which signs what it
594
+ serves; and a **missing** attestation is not a failure, so it does not require
595
+ provenance even though all four packages publish it.
596
+
597
+ > **This step needs a registry that serves `/-/npm/v1/keys`.** If your runner
598
+ > points npm at a mirror or proxy that does not (via `actions/setup-node`'s
599
+ > `registry-url:`, a corporate `~/.npmrc`, or `npm_config_registry`), the
600
+ > install succeeds and this step then fails with `EMISSINGSIGNATUREKEY`. A
601
+ > sigstore outage has the same effect. It fails closed on purpose, so that is a
602
+ > red gate rather than a skipped check; pin to `@v0.4.0`, which does not
603
+ > verify, if it blocks you.
604
+
605
+ `--base` is passed only when the `base-ref` input names one; left empty, the
606
+ umbrella reads `GITHUB_BASE_REF` itself and treats an empty value as "not a
607
+ pull request", which is what a push build wants.
500
608
 
501
609
  On a `pull_request` event the action passes
502
610
  **`--trust-base origin/$GITHUB_BASE_REF`** of its own accord, so the run takes
@@ -513,6 +621,49 @@ event the workflow file itself runs from the pull request's own ref, so an
513
621
  opt-out here would be settable by the very pull request the mode exists to
514
622
  judge: the knob and the thing it protects against would be the same file.
515
623
 
624
+ **The normal way to use this is by tag, not by path.** `uses: ./` reads
625
+ `action.yml` out of whichever tree the workflow runs against, which is
626
+ correct only for this repository's own workflows testing themselves; on a
627
+ pull request from anywhere else it would read `action.yml` out of the pull
628
+ request being judged, and the version-pin protection described above
629
+ "protects nothing" against a tree that controls its own judge (the
630
+ validate step's own comment in `action.yml` says so in those words). Name
631
+ this action by owner and tag instead:
632
+
633
+ ```yaml
634
+ name: guardrails
635
+ on: pull_request
636
+
637
+ jobs:
638
+ gates:
639
+ runs-on: ubuntu-latest
640
+ permissions:
641
+ contents: read
642
+ security-events: write
643
+ steps:
644
+ - uses: actions/checkout@v4
645
+ with:
646
+ fetch-depth: 0
647
+ - uses: actions/setup-node@v4
648
+ with:
649
+ node-version: '22.11.0'
650
+ - id: conductor
651
+ uses: vaultcompasshq/conductor@v0.4.5
652
+ with:
653
+ output: conductor.sarif
654
+ - uses: github/codeql-action/upload-sarif@v3
655
+ if: always()
656
+ continue-on-error: true
657
+ with:
658
+ sarif_file: ${{ steps.conductor.outputs.sarif }}
659
+ ```
660
+
661
+ No version inputs there at all: left out, the four gates run at the versions
662
+ this tag ships, which is the intended default. The fully spelled-out example
663
+ below adds explicit version pins and comments explaining each one, for a
664
+ workflow that wants that transparency; the four lines can still be left out
665
+ entirely to take the tag's own versions, exactly as the form above does.
666
+
516
667
  ```yaml
517
668
  name: guardrails
518
669
  on: pull_request
@@ -535,13 +686,32 @@ jobs:
535
686
  - uses: pnpm/action-setup@v4
536
687
  - uses: actions/setup-node@v4
537
688
  with:
538
- node-version: 22
689
+ # Not a bare major: Node 22.0.0 ships npm 10.5.1, which the action
690
+ # refuses because that client reports a clean install as tampered
691
+ # with. 22.1.0 or later, or 20.13.0 or later, carries an npm that
692
+ # can verify.
693
+ node-version: '22.11.0'
539
694
  cache: pnpm
695
+ # Your own dependencies. The gates are NOT among the things this has to
696
+ # install: the action installs those itself, globally, at the versions
697
+ # pinned below.
540
698
  - run: pnpm install --frozen-lockfile
541
699
  - id: conductor
542
- uses: ./
700
+ uses: vaultcompasshq/conductor@v0.4.5
543
701
  with:
544
702
  output: conductor.sarif
703
+ # Exact versions, never a range and never "latest". These four
704
+ # lines decide which programs judge a pull request, so a pull
705
+ # request's own lockfile no longer does. They are only as protected
706
+ # as this file is: require review on .github/workflows in your
707
+ # branch protection. Bump them like any other pin, in a pull
708
+ # request of their own. Forward only on a pull request: the action
709
+ # refuses a pin below what its tag ships, and the four lines can be
710
+ # left out entirely to take that tag's own versions.
711
+ conductor-version: 0.4.5
712
+ dep-guard-version: 0.7.0
713
+ vault-guard-version: 1.8.0
714
+ intent-guard-version: 1.5.2
545
715
  - uses: github/codeql-action/upload-sarif@v3
546
716
  # Always: the log is most worth having on the run that failed.
547
717
  if: always()
@@ -564,6 +734,99 @@ workflows do.
564
734
 
565
735
  ### The report as a pull request comment
566
736
 
737
+ **Built in, opt-in.** This matters most for an advisory job: a `pull_request`
738
+ check that is not marked required does not block anything when it fails, and
739
+ without a comment its findings live only in the job's exit code and log,
740
+ which teaches a developer nothing. Set `pr-comment: true` on the Action's own
741
+ step and add `pull-requests: write` to the job's `permissions`:
742
+
743
+ ```yaml
744
+ permissions:
745
+ contents: read
746
+ security-events: write
747
+ pull-requests: write
748
+ steps:
749
+ # ... checkout, pnpm, setup-node, install, as in the example above ...
750
+ - id: conductor
751
+ uses: vaultcompasshq/conductor@v0.4.5
752
+ with:
753
+ output: conductor.sarif
754
+ pr-comment: true
755
+ ```
756
+
757
+ That posts conductor's own text report, the one the README quotes above
758
+ ("conductor: clean, nothing blocked. 2 gate(s) ran: ..."), as a comment on
759
+ the pull request. It is **sticky**: a hidden marker in the comment body lets
760
+ a re-run find and update that same comment, so a push does not pile up a new
761
+ comment every time, the way the manual recipe below does. It runs on a
762
+ `pull_request` or `pull_request_target` event only, and it runs whether the
763
+ gates step passed or failed, since a blocking run is the one an advisory
764
+ check most needs a developer to actually see.
765
+
766
+ **It is a no-op on a pull request from a fork.** The default `GITHUB_TOKEN`
767
+ there is read-only regardless of the `pull-requests: write` permission you
768
+ grant, so the post fails; the step catches that, prints a `::warning::`
769
+ naming the likely cause, and continues. The gate's own pass/fail is decided
770
+ entirely by the earlier "Run the gates" step and never depends on whether
771
+ the comment posted, on a fork or anywhere else. `pull_request_target` runs
772
+ with a writable token and the base repository's own workflow instead, which
773
+ is a different security decision to take on purpose rather than a flag to
774
+ add; nothing here does that for you.
775
+
776
+ Off by default, so an existing consumer of this action is unaffected.
777
+
778
+ **The sticky match only ever adopts conductor's own comment.** On
779
+ `pull_request_target` this step runs with a write token even though the
780
+ pull request itself is untrusted, so anyone who can comment on the pull
781
+ request could, in principle, author a comment carrying the same hidden
782
+ marker before conductor's first run. The match requires the marker AND that
783
+ the comment was authored by the GitHub Actions bot; a marker in anyone
784
+ else's comment is never adopted, and conductor creates its own comment
785
+ instead, exactly as if no marked comment existed.
786
+
787
+ **Running this Action more than once against the same pull request** (for
788
+ example, once per package in a monorepo) needs its own marker per run, or
789
+ those runs fight over one shared comment. Set `pr-comment-marker` to a
790
+ distinct string per invocation and each run stays sticky to its own
791
+ comment; left unset (the default), every invocation uses the same built-in
792
+ marker, which is the existing, unchanged behaviour for a single invocation
793
+ per pull request.
794
+
795
+ ```yaml
796
+ - id: conductor-package-a
797
+ uses: vaultcompasshq/conductor@v0.4.5
798
+ with:
799
+ pr-comment: true
800
+ pr-comment-marker: 'package-a'
801
+ - id: conductor-package-b
802
+ uses: vaultcompasshq/conductor@v0.4.5
803
+ with:
804
+ pr-comment: true
805
+ pr-comment-marker: 'package-b'
806
+ ```
807
+
808
+ **A note for anyone changing this surface, not for a consumer of the
809
+ Action**: the comment body reaches `gh` with `-F body=@<file>` (`gh api`'s
810
+ file-read form), never `-f`, which sends the value as a literal string
811
+ instead of reading the file. The offline test suite
812
+ (`scripts/tests/pr-comment.test.mjs`, `scripts/tests/pr-comment-cli.test.mjs`)
813
+ replaces `gh` with a recorder or a shim, and neither can tell `-f` from `-F`
814
+ apart, since both simply accept a `key=value` string and neither actually
815
+ reads the `@file` form. `scripts/tests/pr-comment-smoke.test.mjs` is what
816
+ actually proves the distinction, against a real `gh` binary; it is skipped
817
+ by an ordinary offline `pnpm test` run and only runs with `GH_TOKEN` (or
818
+ `GITHUB_TOKEN`), `CONDUCTOR_PR_COMMENT_SMOKE_REPO`, and
819
+ `CONDUCTOR_PR_COMMENT_SMOKE_ISSUE` set. Run it for real against a scratch
820
+ issue or pull request before any release that touches
821
+ `scripts/lib/pr-comment.mjs` or `scripts/pr-comment.mjs`; the file itself
822
+ documents the exact invocation.
823
+
824
+ **The manual recipe below still has a reason to exist**: a non-sticky
825
+ comment (one per run, never edited), a report you want to post yourself with
826
+ different formatting, or a workflow that would rather not add
827
+ `pull-requests: write` to the same job the gates run in. For the common
828
+ case, `pr-comment: true` is the built-in answer.
829
+
567
830
  The SARIF upload produces no alerts on a private repository without GitHub
568
831
  Code Security, which is why that step carries `continue-on-error`. A comment
569
832
  is free there. Add `pull-requests: write` to the job's `permissions` and this
@@ -580,19 +843,41 @@ writes SARIF and nothing else, so the job costs roughly twice the gate time;
580
843
  env:
581
844
  GH_TOKEN: ${{ github.token }}
582
845
  PR: ${{ github.event.pull_request.number }}
846
+ # The same derivation the Action makes for its own --trust-base.
847
+ # github.base_ref is a BRANCH NAME and is empty outside a pull
848
+ # request, so an empty value has to mean "not a pull request":
849
+ # passing --trust-base unconditionally would hand a push build the
850
+ # literal ref "origin/" and fail it closed.
851
+ BASE: ${{ github.base_ref }}
583
852
  # || true: a blocking finding is a non-zero exit, and that verdict
584
853
  # belongs to the gates step rather than to this one.
585
854
  run: |
586
- node_modules/.bin/conductor run --stage ci --format text --verbose --output conductor.txt || true
855
+ ARGS=(run --stage ci --format text --verbose --output conductor.txt)
856
+ if [ -n "$BASE" ]; then
857
+ ARGS+=(--trust-base "origin/$BASE")
858
+ fi
859
+ conductor "${ARGS[@]}" || true
587
860
  gh pr comment "$PR" --body-file conductor.txt
588
861
  ```
589
862
 
863
+ `conductor` and not `node_modules/.bin/conductor`: the gates step put the
864
+ installed one on `PATH`. Reaching into `node_modules` here would report a
865
+ different program's verdict from the one in the uploaded log.
866
+
867
+ `--trust-base` for the same reason, and the two together are what make this
868
+ the run the uploaded log is about: same binary AND same trust base. Without
869
+ it this second run reads the head's own policy file, consults
870
+ `node_modules/.bin` like any local run, and never asks where a gate's program
871
+ came from, so a pull request that plants one gets a clean comment beside a
872
+ SARIF log that refused it.
873
+
590
874
  **Mirror whatever you gave the Action**, or the two runs can report different
591
875
  contracts. The step above matches the example, which passes neither
592
- `base-ref` nor `spec`, so both runs read `GITHUB_BASE_REF` and the `Spec:`
593
- line out of the job environment themselves and land on the same contract. If
594
- you set either input on the gates step, pass the same values here as `--base`
595
- and `--spec`; if you do not, the comment is a report of a contract source the
876
+ `base-ref`, `trust-base` nor `spec`, so both runs read `GITHUB_BASE_REF` and
877
+ the `Spec:` line out of the job environment themselves and land on the same
878
+ contract and the same ref. If you set any of those inputs on the gates step,
879
+ pass the same values here as `--base`, `--trust-base` and `--spec`; if you do
880
+ not, the comment is a report of a contract source, or of a boundary, the
596
881
  uploaded log never used.
597
882
 
598
883
  **On a pull request from a fork the token is read-only**, so `gh pr comment`
package/dist/cli.js CHANGED
@@ -79,49 +79,89 @@ function repoRoot(cwd) {
79
79
  * reported, and, when the trust base itself cannot be used, it names the
80
80
  * gates the report says did not run.
81
81
  */
82
+ /**
83
+ * An inventory, and only an inventory: a list of gate NAMES for the report,
84
+ * taken from the one file available when the trust base itself cannot supply
85
+ * a policy. It is head-controlled, so it can be SHORTER as well as longer
86
+ * than the base's -- every gate `enabled: false`, or a file that will not
87
+ * parse at all, leaves it empty.
88
+ *
89
+ * Neither direction can weaken the verdict, and that is the property this
90
+ * rests on rather than on the inventory being right. refusedTrustBase
91
+ * enforces every gate it names and writes exit 2 itself, and the refusal is
92
+ * carried on the result so both renderers lead with it whether the inventory
93
+ * names three gates or none. A longer list makes the report longer; a
94
+ * shorter one makes it shorter; the verdict is the same sentence either way.
95
+ *
96
+ * An unreadable head file leaves the list empty rather than throwing: the
97
+ * trust base is what went wrong, and reporting the head's malformed file
98
+ * would send the reader to the wrong fix on a run that would have ignored
99
+ * that file anyway.
100
+ */
101
+ function inventoryFromHead(root, overrides) {
102
+ try {
103
+ return applyCliOverrides(loadPolicy(root), overrides);
104
+ }
105
+ catch {
106
+ return { version: 1, gates: {}, report: { format: 'text' } };
107
+ }
108
+ }
82
109
  function policyForRun(root, trustBase, overrides) {
83
110
  if (trustBase === undefined) {
84
111
  return { kind: 'policy', policy: applyCliOverrides(loadPolicy(root), overrides) };
85
112
  }
86
113
  const refusal = refuseTrustBaseRef(root, trustBase);
87
114
  if (refusal !== null) {
88
- // An inventory, and only an inventory: a list of gate NAMES for the
89
- // report, taken from the one file available when the base ref cannot be
90
- // read. It is head-controlled, so it can be SHORTER as well as longer
91
- // than the base's -- every gate `enabled: false`, or a file that will not
92
- // parse at all, leaves it empty -- and an earlier comment here claimed
93
- // only the longer direction, which is how the empty case went unnoticed.
94
- //
95
- // Neither direction can weaken the verdict, and that is the property this
96
- // rests on rather than on the inventory being right. refusedTrustBase
97
- // enforces every gate it names and writes exit 2 itself, and the refusal
98
- // is carried on the result so both renderers lead with it whether the
99
- // inventory names three gates or none. A longer list makes the report
100
- // longer; a shorter one makes it shorter; the verdict is the same
101
- // sentence either way.
102
- //
103
- // An unreadable head file leaves the list empty rather than replacing the
104
- // refusal with a policy error: the ref is what went wrong, and reporting
105
- // the head's malformed file would send the reader to the wrong fix on a
106
- // run that would have ignored that file anyway.
107
- let inventory = { version: 1, gates: {}, report: { format: 'text' } };
108
- try {
109
- inventory = applyCliOverrides(loadPolicy(root), overrides);
110
- }
111
- catch {
112
- inventory = { version: 1, gates: {}, report: { format: 'text' } };
113
- }
114
- return { kind: 'refused', ref: trustBase, detail: refusal, inventory };
115
+ return {
116
+ kind: 'refused',
117
+ ref: trustBase,
118
+ detail: refusal,
119
+ inventory: inventoryFromHead(root, overrides),
120
+ };
115
121
  }
116
122
  const baseText = readPolicyAtRef(root, trustBase);
117
123
  const headText = readPolicyAtRef(root, 'HEAD');
118
124
  if (baseText === null) {
119
- throw new PolicyError(`No ${POLICY_FILE_NAME} on "${trustBase}". On a pull-request run every rule comes from the ` +
125
+ // CONFIG ABSENT ON THE BASE, not a policy error thrown from here. The old
126
+ // shape threw a PolicyError, which the run command's catch prints as one
127
+ // line on STDERR and nothing else: no report on stdout, no SARIF file
128
+ // written, exit code 2 by accident of `fail()` rather than by the same
129
+ // could-not-run machinery every other trust-base failure goes through.
130
+ // Two real adopters, running in advisory mode, hit exactly this on their
131
+ // first pull request -- before "conductor init" had ever landed on their
132
+ // base branch -- and read a red step with an empty report as either
133
+ // nothing to see or a broken tool, never as "finish adopting".
134
+ //
135
+ // This is DISCOVERED-empty, not IMPOSED-empty: nobody wrote a policy file
136
+ // down and disabled every gate in it, the base ref simply has none yet.
137
+ // That is the same "nothing here is a result of any kind" shape as a base
138
+ // ref that will not resolve, so it goes through the exact same
139
+ // refusedTrustBase reporting: could-not-run (exit 2), the reason on the
140
+ // text report and in the SARIF log, and the head's own gates named as not
141
+ // having run. A base ref whose file is merely ABSENT must not be treated
142
+ // more gently than one that will not resolve at all; both mean this run
143
+ // has no policy to have been judged by.
144
+ //
145
+ // CONTRAST: a base ref that DOES carry a policy file in which the user
146
+ // switched every gate off, or deferred them all past this stage, is left
147
+ // alone below in the ordinary `runAll` path and keeps reporting a clean
148
+ // exit 0 -- see the "no gate ran because none is enabled" verdict in
149
+ // output-text.ts. That is an IMPOSED-empty state, a decision written down
150
+ // on the base branch on purpose, and mirrors the no-contract
151
+ // (discovered, write one) vs contract-waived (imposed, a person already
152
+ // decided) distinction the intent gate already makes for a missing spec.
153
+ const detail = `No ${POLICY_FILE_NAME} on "${trustBase}". On a pull-request run every rule comes from the ` +
120
154
  'base ref, so this run has no policy at all and nothing was checked. ' +
121
155
  (headText === null
122
156
  ? `Run "conductor init" on the base branch.`
123
157
  : `The ${POLICY_FILE_NAME} in this pull request is a proposal: it decides what runs ` +
124
- 'once it is on the base branch, and never on the pull request that adds it.'));
158
+ 'once it is on the base branch, and never on the pull request that adds it.');
159
+ return {
160
+ kind: 'refused',
161
+ ref: trustBase,
162
+ detail,
163
+ inventory: inventoryFromHead(root, overrides),
164
+ };
125
165
  }
126
166
  return {
127
167
  kind: 'policy',
@@ -141,6 +141,17 @@ export interface GateOutcome {
141
141
  * pull-request mode, and what it said was proposed.
142
142
  */
143
143
  trustBase?: GateTrustBase;
144
+ /**
145
+ * The `node_modules/.bin` candidate resolution did NOT try, repository
146
+ * relative, on a pull-request run where one was there.
147
+ *
148
+ * Absent outside pull-request mode, and absent inside it when there was
149
+ * nothing to skip: it is a claim that something was there and was declined,
150
+ * not a claim about the mode. Reported so an adopter whose gates are
151
+ * devDependencies learns why the gate came from somewhere else, or from
152
+ * nowhere. Repository-relative because it reaches a published log.
153
+ */
154
+ nodeModulesSkipped?: string;
144
155
  }
145
156
  export interface RunGateOptions {
146
157
  repoRoot: string;
@@ -178,6 +189,37 @@ export interface RunGateOptions {
178
189
  * would have to learn about.
179
190
  */
180
191
  export declare function preparationFailed(gate: GatePolicy, detail: string): GateOutcome;
192
+ /**
193
+ * EVERY program this binary would spawn, vetted against the trust base.
194
+ *
195
+ * Two files and not one. `program` is what the gate runs, and
196
+ * `versionProbe.command` can be a DIFFERENT file: for a per-command binary
197
+ * that ignores `--version`, resolution finds the version-safe sibling
198
+ * somewhere else entirely (`versionProbeFor`, src/resolve.ts:238-262). Vetting
199
+ * only the first left the probe free to spawn whatever the head put at the
200
+ * second, and a probe RUNS a program as thoroughly as a scan does.
201
+ *
202
+ * The first refusal wins and the rest is not consulted. The verdict is the
203
+ * same either way, and the sentence a reader has to act on should name one
204
+ * file rather than two.
205
+ */
206
+ export declare function refuseHeadControlledBinary(repoRoot: string, trustBase: string, binary: ResolvedBinary): string | null;
207
+ /**
208
+ * The outcome for a gate whose program the umbrella declined to run.
209
+ *
210
+ * ONE BUILDER, TWO CALLERS, and that is the point of it being here rather
211
+ * than written out at the return that needs it. `runGate` reaches this before
212
+ * it spawns anything; `runAll` reaches it earlier still, before the intent
213
+ * gate's preparation spawns the same program three times. A second copy of
214
+ * the shape would be a second chance to forget the override below.
215
+ *
216
+ * ENFORCED, whatever the policy says. `enforce: false` is a standing decision
217
+ * about what a gate's FINDINGS are worth, and this gate produced none: the
218
+ * umbrella refused to run a program the pull request chose. Letting an
219
+ * unenforced gate swallow that would let a pull request pick its own judge
220
+ * and keep the run green.
221
+ */
222
+ export declare function gateProgramRefused(gate: GatePolicy, trustBase: string, refusal: string, binary: ResolvedBinary): GateOutcome;
181
223
  /**
182
224
  * The arguments the umbrella adds, per product.
183
225
  *
@@ -20,7 +20,7 @@ import { NormalizeError } from './envelope.js';
20
20
  import { normalizeDepGuard, normalizeFailedGate, normalizeIntentGuard, normalizeMisconfiguredGate, normalizeMissingGate, normalizeUnparseableGate, normalizeVaultGuard, } from './normalize.js';
21
21
  import { renderOptionFlags } from './policy.js';
22
22
  import { atLeastVersion, refuseHeadControlledProgram } from './trust-base.js';
23
- import { ResolveError, candidateNames, resolveGateBinary } from './resolve.js';
23
+ import { ResolveError, candidateNames, nodeModulesCandidate, resolveGateBinary, } from './resolve.js';
24
24
  /**
25
25
  * The first version of each gate that understands `--trust-base`.
26
26
  *
@@ -142,6 +142,66 @@ const EMPTY_RUN = {
142
142
  diagnostics: [],
143
143
  details: {},
144
144
  };
145
+ /**
146
+ * EVERY program this binary would spawn, vetted against the trust base.
147
+ *
148
+ * Two files and not one. `program` is what the gate runs, and
149
+ * `versionProbe.command` can be a DIFFERENT file: for a per-command binary
150
+ * that ignores `--version`, resolution finds the version-safe sibling
151
+ * somewhere else entirely (`versionProbeFor`, src/resolve.ts:238-262). Vetting
152
+ * only the first left the probe free to spawn whatever the head put at the
153
+ * second, and a probe RUNS a program as thoroughly as a scan does.
154
+ *
155
+ * The first refusal wins and the rest is not consulted. The verdict is the
156
+ * same either way, and the sentence a reader has to act on should name one
157
+ * file rather than two.
158
+ */
159
+ export function refuseHeadControlledBinary(repoRoot, trustBase, binary) {
160
+ for (const program of [binary.program, binary.versionProbe?.command]) {
161
+ if (program === undefined) {
162
+ continue;
163
+ }
164
+ const refusal = refuseHeadControlledProgram(repoRoot, trustBase, program);
165
+ if (refusal !== null) {
166
+ return refusal;
167
+ }
168
+ }
169
+ return null;
170
+ }
171
+ /**
172
+ * The outcome for a gate whose program the umbrella declined to run.
173
+ *
174
+ * ONE BUILDER, TWO CALLERS, and that is the point of it being here rather
175
+ * than written out at the return that needs it. `runGate` reaches this before
176
+ * it spawns anything; `runAll` reaches it earlier still, before the intent
177
+ * gate's preparation spawns the same program three times. A second copy of
178
+ * the shape would be a second chance to forget the override below.
179
+ *
180
+ * ENFORCED, whatever the policy says. `enforce: false` is a standing decision
181
+ * about what a gate's FINDINGS are worth, and this gate produced none: the
182
+ * umbrella refused to run a program the pull request chose. Letting an
183
+ * unenforced gate swallow that would let a pull request pick its own judge
184
+ * and keep the run green.
185
+ */
186
+ export function gateProgramRefused(gate, trustBase, refusal, binary) {
187
+ return {
188
+ role: gate.role,
189
+ product: gate.product,
190
+ stage: gate.stage,
191
+ enforce: true,
192
+ productVersion: null,
193
+ argv: [],
194
+ binary,
195
+ exitCode: null,
196
+ durationMs: 0,
197
+ stderr: '',
198
+ trustBase: { ref: trustBase, withheld: null, refused: refusal, proposals: [] },
199
+ couldNotRun: { reason: 'gate-program-refused', detail: refusal },
200
+ findings: [normalizeFailedGate(gate.role, gate.product, refusal)],
201
+ run: EMPTY_RUN,
202
+ diagnostics: [],
203
+ };
204
+ }
145
205
  /**
146
206
  * The arguments the umbrella adds, per product.
147
207
  *
@@ -267,6 +327,31 @@ function messageOf(err) {
267
327
  }
268
328
  return String(err);
269
329
  }
330
+ /**
331
+ * The sentence a gate that resolved to nothing gets on a pull-request run.
332
+ *
333
+ * EMPTY OUTSIDE PULL-REQUEST MODE, because outside it nothing about
334
+ * resolution changed and the old message is still the whole story. Inside it
335
+ * the message would otherwise be actively misleading in the common case: an
336
+ * adopter with the gates as devDependencies is looking straight at a
337
+ * node_modules/.bin/<gate> while being told no binary was found.
338
+ *
339
+ * It names `npm install -g` and the Action's own input rather than "install
340
+ * it", because those are the two places the fix actually is: on a runner, the
341
+ * Action installs the gates itself at a version pinned in a workflow file that
342
+ * lives on the base branch, which is the protected side.
343
+ */
344
+ function missingGateRemedy(product, skipNodeModules, skipped) {
345
+ if (!skipNodeModules) {
346
+ return '';
347
+ }
348
+ return (' On a pull-request run node_modules/.bin is not consulted at all: what is installed there ' +
349
+ 'is chosen by the head own manifest and lockfile, so no ref approves it.' +
350
+ (skipped === null ? '' : ` There is a ${skipped}, and it was skipped for that reason.`) +
351
+ ` Install the gate outside the tree with npm install -g @vaultcompass/${product}. The ` +
352
+ `conductor Action does exactly that, at the version its ${product}-version input pins, and ` +
353
+ 'that pin lives in the workflow file on the base branch.');
354
+ }
270
355
  /**
271
356
  * Runs one gate. TOTAL: this never throws.
272
357
  *
@@ -337,10 +422,20 @@ export function runGate(gate, options) {
337
422
  }
338
423
  function runGateInner(gate, options, started, progress) {
339
424
  const timeoutMs = options.timeoutMs ?? 120_000;
425
+ // ONE PLACE DECIDES WHAT PULL-REQUEST MODE IS, and it is the presence of a
426
+ // trust base. resolve.ts knows nothing about refs, so it is told rather than
427
+ // asked.
428
+ const skipNodeModules = options.trustBase !== undefined;
429
+ const skipped = skipNodeModules ? nodeModulesCandidate(gate, options.repoRoot) : null;
430
+ if (skipped !== null) {
431
+ // Onto `progress` rather than only onto the returns below, so the
432
+ // backstop's view carries it too.
433
+ Object.assign(progress, { nodeModulesSkipped: skipped });
434
+ }
340
435
  const base = progress;
341
436
  let binary;
342
437
  try {
343
- binary = resolveGateBinary(gate, options.repoRoot, options.pathValue);
438
+ binary = resolveGateBinary(gate, options.repoRoot, options.pathValue, { skipNodeModules });
344
439
  }
345
440
  catch (err) {
346
441
  // A ResolveError is the expected shape here; anything else is still a
@@ -364,6 +459,7 @@ function runGateInner(gate, options, started, progress) {
364
459
  };
365
460
  }
366
461
  if (binary === null) {
462
+ const remedy = missingGateRemedy(gate.product, skipNodeModules, skipped);
367
463
  return {
368
464
  ...base,
369
465
  durationMs: Date.now() - started,
@@ -371,13 +467,19 @@ function runGateInner(gate, options, started, progress) {
371
467
  reason: 'binary-missing',
372
468
  // Named in resolution order, which is the repository's own copy
373
469
  // first. Saying PATH first points a reader at the location this
374
- // tool prefers second, which is the wrong place to install it.
375
- detail: `no ${gate.product} binary in node_modules/.bin or on PATH`,
470
+ // tool prefers second, which is the wrong place to install it. On a
471
+ // pull-request run that order has one entry, because the other one is
472
+ // not a location this run has.
473
+ detail: (skipNodeModules
474
+ ? `no ${gate.product} binary on PATH`
475
+ : `no ${gate.product} binary in node_modules/.bin or on PATH`) + remedy,
376
476
  },
377
477
  // A missing enabled gate is a finding of the umbrella's own, never a
378
478
  // silent skip. Skipping is how a gate ends up switched on in the
379
479
  // policy file and absent in reality for months.
380
- findings: [normalizeMissingGate(gate.role, gate.product, candidateNames(gate.product))],
480
+ findings: [
481
+ normalizeMissingGate(gate.role, gate.product, candidateNames(gate.product), remedy),
482
+ ],
381
483
  run: EMPTY_RUN,
382
484
  diagnostics: [],
383
485
  };
@@ -388,23 +490,16 @@ function runGateInner(gate, options, started, progress) {
388
490
  // version of this attack and it would have been run before anything checked
389
491
  // where it came from.
390
492
  if (options.trustBase !== undefined) {
391
- const refusal = refuseHeadControlledProgram(options.repoRoot, options.trustBase, binary.program);
493
+ // BOTH SPAWNABLE PATHS, not just the program: the probe below can be a
494
+ // different file from binary.program, and it is executed just as
495
+ // thoroughly. The enforce override and the rest of the shape live in
496
+ // gateProgramRefused, which the intent gate's preparation shares.
497
+ const refusal = refuseHeadControlledBinary(options.repoRoot, options.trustBase, binary);
392
498
  if (refusal !== null) {
393
499
  return {
394
500
  ...base,
395
- binary,
501
+ ...gateProgramRefused(gate, options.trustBase, refusal, binary),
396
502
  durationMs: Date.now() - started,
397
- // ENFORCED, whatever the policy says. enforce: false is a standing
398
- // decision about what a gate's FINDINGS are worth, and this gate
399
- // produced none: the umbrella refused to run a program the pull
400
- // request chose. Letting an unenforced gate swallow that would let a
401
- // pull request pick its own judge and keep the run green.
402
- enforce: true,
403
- trustBase: { ref: options.trustBase, withheld: null, refused: refusal, proposals: [] },
404
- couldNotRun: { reason: 'gate-program-refused', detail: refusal },
405
- findings: [normalizeFailedGate(gate.role, gate.product, refusal)],
406
- run: EMPTY_RUN,
407
- diagnostics: [],
408
503
  };
409
504
  }
410
505
  }
@@ -45,7 +45,18 @@ export type GateProblemRule = 'conductor/gate-missing' | 'conductor/gate-output-
45
45
  * looked. So it is a finding, it blocks, and it says which names were
46
46
  * tried, so the fix is in the message rather than in a support thread.
47
47
  */
48
- export declare function normalizeMissingGate(role: GateRole, product: Product, candidates: string[]): Finding;
48
+ export declare function normalizeMissingGate(role: GateRole, product: Product, candidates: string[],
49
+ /**
50
+ * A sentence appended to the message, or nothing.
51
+ *
52
+ * Written by the caller rather than decided here, because the only thing
53
+ * that varies is WHY the search came up empty, and that is a fact about the
54
+ * run rather than about the finding. Today the one caller that passes
55
+ * anything is a pull-request run, where node_modules/.bin was never
56
+ * searched and the message would otherwise be misleading in exactly the
57
+ * repository shape this is most likely to happen in.
58
+ */
59
+ remedy?: string): Finding;
49
60
  /**
50
61
  * The finding raised when a gate's `command:` names a file that is not
51
62
  * there.
package/dist/normalize.js CHANGED
@@ -620,10 +620,21 @@ function gateProblem(ruleId, role, product, message, details) {
620
620
  * looked. So it is a finding, it blocks, and it says which names were
621
621
  * tried, so the fix is in the message rather than in a support thread.
622
622
  */
623
- export function normalizeMissingGate(role, product, candidates) {
623
+ export function normalizeMissingGate(role, product, candidates,
624
+ /**
625
+ * A sentence appended to the message, or nothing.
626
+ *
627
+ * Written by the caller rather than decided here, because the only thing
628
+ * that varies is WHY the search came up empty, and that is a fact about the
629
+ * run rather than about the finding. Today the one caller that passes
630
+ * anything is a pull-request run, where node_modules/.bin was never
631
+ * searched and the message would otherwise be misleading in exactly the
632
+ * repository shape this is most likely to happen in.
633
+ */
634
+ remedy = '') {
624
635
  return gateProblem('conductor/gate-missing', role, product, `The "${role}" gate is enabled but no ${product} binary was found. ` +
625
636
  `Looked for: ${candidates.join(', ')}. Install it, point the gate at a build with an ` +
626
- `absolute "command:", or set enabled: false to switch the gate off on purpose.`, { candidates });
637
+ `absolute "command:", or set enabled: false to switch the gate off on purpose.${remedy}`, { candidates });
627
638
  }
628
639
  /**
629
640
  * The finding raised when a gate's `command:` names a file that is not
@@ -588,10 +588,11 @@ function proposalNotifications(result) {
588
588
  /**
589
589
  * The trust base the umbrella refused, when it refused one.
590
590
  *
591
- * ERROR LEVEL, and the only notification here that is not a note. Every other
592
- * one says how much of the policy a run covered; this one says the run did
593
- * not happen. It is the one statement in this file that a reader must not be
594
- * able to scroll past.
591
+ * ERROR LEVEL, one of the two notifications here that are not notes, and the
592
+ * wider of the two: a refused program says one gate did not run, and this one
593
+ * says the whole run did not. Every other notification in this file says how
594
+ * much of the policy a run covered. These two say what was not covered at
595
+ * all, and a reader must not be able to scroll past either.
595
596
  *
596
597
  * A NOTIFICATION rather than a result even so, and the reason is the
597
598
  * discriminator further down: a result is about a place in somebody's code,
@@ -686,6 +687,40 @@ function trustBaseWithheldNotifications(result) {
686
687
  ];
687
688
  });
688
689
  }
690
+ /**
691
+ * The node_modules/.bin candidates a pull-request run declined to take.
692
+ *
693
+ * ONE NOTIFICATION FOR THE RUN, not one per gate. This is a statement about
694
+ * the run's mode: node_modules/.bin is not a location a pull-request run has,
695
+ * and in the repository shape it happens in most, gates installed as
696
+ * devDependencies and nothing else, it is true of every gate at once.
697
+ *
698
+ * A NOTE, by the discriminator above, and not a close call. Nothing went
699
+ * wrong, nobody's code is at fault, and the gates still ran: it says WHERE the
700
+ * program that ran came from, which is a coverage statement in the same shape
701
+ * as the withheld one. When the skip actually cost a gate its run, that gate's
702
+ * own could-not-run result carries the remedy, so a consumer reading only
703
+ * results still learns what to do.
704
+ */
705
+ function nodeModulesSkippedNotifications(result) {
706
+ const skipped = result.gates.flatMap((gate) => gate.nodeModulesSkipped === undefined
707
+ ? []
708
+ : [{ role: gate.role, product: gate.product, path: gate.nodeModulesSkipped }]);
709
+ if (skipped.length === 0) {
710
+ return [];
711
+ }
712
+ return [
713
+ {
714
+ id: 'conductor/node-modules-skipped',
715
+ message: 'Resolution did NOT consult node_modules/.bin on this pull-request run: ' +
716
+ `${skipped.map((entry) => `${entry.role} (${entry.product}) at ${entry.path}`).join(', ')}. ` +
717
+ 'What is installed there is chosen by the head own manifest and lockfile, so no ref ' +
718
+ 'approves it. Each of those gates was resolved from PATH or from an absolute command: ' +
719
+ 'instead, or reported as could-not-run with the install remedy.',
720
+ details: { ref: result.trustBase?.ref ?? null, skipped },
721
+ },
722
+ ];
723
+ }
689
724
  /**
690
725
  * The gates the policy file told not to decide anything, as notifications.
691
726
  *
@@ -799,6 +834,7 @@ export function renderSarif(result, umbrellaVersion) {
799
834
  ...programRefusedNotifications(result),
800
835
  ...proposalNotifications(result),
801
836
  ...trustBaseWithheldNotifications(result),
837
+ ...nodeModulesSkippedNotifications(result),
802
838
  ];
803
839
  // Notifications earn the run on their own. Before this, the umbrella run
804
840
  // existed only when it had findings and a deferred gate was one; without
@@ -245,12 +245,35 @@ function withheldTrustBase(result) {
245
245
  return reason === undefined || reason === null ? [] : [{ gate, reason }];
246
246
  });
247
247
  }
248
+ /**
249
+ * The gates whose own node_modules/.bin copy this run declined to take.
250
+ *
251
+ * ONE LINE FOR THE WHOLE RUN rather than one per gate. The fact is about the
252
+ * run's mode, not about any gate's verdict, and in the repository shape it is
253
+ * most likely to happen in, gates installed as devDependencies and nothing
254
+ * else, it is true of all three at once. Three lines saying the same thing
255
+ * would read as three problems.
256
+ */
257
+ function nodeModulesSkippedLine(result) {
258
+ const skipped = result.gates.flatMap((gate) => gate.nodeModulesSkipped === undefined
259
+ ? []
260
+ : [`${gate.role} (${gate.product}) at ${gate.nodeModulesSkipped}`]);
261
+ if (skipped.length === 0) {
262
+ return [];
263
+ }
264
+ return [
265
+ ` node_modules/.bin not consulted: ${skipped.join(', ')}. What is installed there is ` +
266
+ 'chosen by the head own manifest and lockfile, so no ref approves it; each of those gates ' +
267
+ 'was resolved from PATH or from an absolute command: instead, or reported as could-not-run.',
268
+ ];
269
+ }
248
270
  /**
249
271
  * The pull-request-mode block of the full report.
250
272
  *
251
- * Three things in one place, because they answer one question between them:
273
+ * Four things in one place, because they answer one question between them:
252
274
  * where the rules came from, what this pull request proposes to change them
253
- * to, and which gates were not covered by any of it.
275
+ * to, which gates were not covered by any of it, and where the programs that
276
+ * did run came from.
254
277
  */
255
278
  function trustBaseLines(result) {
256
279
  if (result.trustBase === null) {
@@ -265,6 +288,7 @@ function trustBaseLines(result) {
265
288
  for (const { gate, reason } of withheldTrustBase(result)) {
266
289
  lines.push(` NOT in pull-request mode ${gate.role} ${gate.product} ${reason}`);
267
290
  }
291
+ lines.push(...nodeModulesSkippedLine(result));
268
292
  return lines;
269
293
  }
270
294
  /**
package/dist/resolve.d.ts CHANGED
@@ -41,9 +41,36 @@ export interface ResolvedBinary {
41
41
  versionProbe: VersionProbe | null;
42
42
  }
43
43
  export declare const CANDIDATES: Record<Product, Candidate[]>;
44
+ export interface ResolveOptions {
45
+ /**
46
+ * Leave `node_modules/.bin` out of the search entirely.
47
+ *
48
+ * True on a pull-request run and false everywhere else, decided by the
49
+ * caller from whether a trust base is set rather than re-derived here: this
50
+ * module knows nothing about refs, and there must be exactly one place that
51
+ * says what pull-request mode is.
52
+ */
53
+ skipNodeModules?: boolean;
54
+ }
44
55
  export declare class ResolveError extends Error {
45
56
  constructor(message: string);
46
57
  }
58
+ /**
59
+ * The node_modules/.bin candidate a pull-request run declined to take.
60
+ *
61
+ * Repository-relative, because it goes into a report and a published SARIF
62
+ * log, where an absolute path is somebody's machine layout.
63
+ *
64
+ * Returns null when the policy names a `command:`: that overrides resolution
65
+ * entirely, so no candidate was skipped, and saying one was would send a
66
+ * reader to fix a file that decided nothing.
67
+ *
68
+ * Separate from `resolveGateBinary` rather than a field on its result,
69
+ * because the case that most needs reporting is the one where resolution
70
+ * returns NOTHING: a repository with the gates only as devDependencies, which
71
+ * is the shape this whole change exists for.
72
+ */
73
+ export declare function nodeModulesCandidate(gate: GatePolicy, repoRoot: string): string | null;
47
74
  /**
48
75
  * Resolves the binary for one gate.
49
76
  *
@@ -58,6 +85,6 @@ export declare class ResolveError extends Error {
58
85
  * named one specific file and falling back to something else would run a
59
86
  * different tool than the one they asked for.
60
87
  */
61
- export declare function resolveGateBinary(gate: GatePolicy, repoRoot: string, pathValue: string): ResolvedBinary | null;
88
+ export declare function resolveGateBinary(gate: GatePolicy, repoRoot: string, pathValue: string, options?: ResolveOptions): ResolvedBinary | null;
62
89
  /** Every name resolution would try, for the missing-gate finding's message. */
63
90
  export declare function candidateNames(product: Product): string[];
package/dist/resolve.js CHANGED
@@ -24,6 +24,19 @@
24
24
  // no say in that; a LOCATION is a statement about which build, and the
25
25
  // repository does.
26
26
  //
27
+ // 1c. ON A PULL-REQUEST RUN node_modules/.bin IS NOT A LOCATION AT ALL.
28
+ // Rule 1b says a LOCATION is a statement about which build and the
29
+ // repository gets to make it. On a pull request the repository making
30
+ // that statement IS the pull request: what is under node_modules was
31
+ // chosen by the head's own manifest and lockfile, installed by a step
32
+ // that runs before the gates, and git has no record of those bytes at
33
+ // either ref. 0.3.0 let resolution take it and then refused the program,
34
+ // which was safe and made every ordinary pull request in a
35
+ // devDependency-installed repository exit 2 with three refusals. Not
36
+ // trying that location at all is the same safety with the gate still
37
+ // running, from a build the base branch pinned. The skip is never
38
+ // silent: the caller reports the candidate it declined.
39
+ //
27
40
  // 2. Only a unified binary is asked for its version. `intent-guard`
28
41
  // answers --version; a per-command binary once did not parse the flag at
29
42
  // all and RAN THE GATE against the current directory instead, and a
@@ -104,12 +117,16 @@ function findInNodeModules(name, repoRoot) {
104
117
  const full = path.join(repoRoot, 'node_modules', '.bin', name);
105
118
  return isExecutableFile(full) ? full : null;
106
119
  }
107
- function locate(name, repoRoot, pathValue) {
120
+ function locate(name, repoRoot, pathValue, skipNodeModules) {
108
121
  // The repository's own copy first: a project pin beats a global install,
109
- // which is what "pnpm exec <name>" does in the same repository.
110
- const local = findInNodeModules(name, repoRoot);
111
- if (local !== null) {
112
- return { command: local, source: 'node_modules' };
122
+ // which is what "pnpm exec <name>" does in the same repository. Unless this
123
+ // is a pull-request run, where the repository saying which build to use is
124
+ // the pull request itself (rule 1c at the top of this file).
125
+ if (!skipNodeModules) {
126
+ const local = findInNodeModules(name, repoRoot);
127
+ if (local !== null) {
128
+ return { command: local, source: 'node_modules' };
129
+ }
113
130
  }
114
131
  const onPath = findOnPath(name, pathValue);
115
132
  if (onPath !== null) {
@@ -117,8 +134,34 @@ function locate(name, repoRoot, pathValue) {
117
134
  }
118
135
  return null;
119
136
  }
137
+ /**
138
+ * The node_modules/.bin candidate a pull-request run declined to take.
139
+ *
140
+ * Repository-relative, because it goes into a report and a published SARIF
141
+ * log, where an absolute path is somebody's machine layout.
142
+ *
143
+ * Returns null when the policy names a `command:`: that overrides resolution
144
+ * entirely, so no candidate was skipped, and saying one was would send a
145
+ * reader to fix a file that decided nothing.
146
+ *
147
+ * Separate from `resolveGateBinary` rather than a field on its result,
148
+ * because the case that most needs reporting is the one where resolution
149
+ * returns NOTHING: a repository with the gates only as devDependencies, which
150
+ * is the shape this whole change exists for.
151
+ */
152
+ export function nodeModulesCandidate(gate, repoRoot) {
153
+ if (gate.command !== undefined) {
154
+ return null;
155
+ }
156
+ for (const candidate of CANDIDATES[gate.product]) {
157
+ if (findInNodeModules(candidate.name, repoRoot) !== null) {
158
+ return `node_modules/.bin/${candidate.name}`;
159
+ }
160
+ }
161
+ return null;
162
+ }
120
163
  /** The candidate whose --version can be asked without running anything. */
121
- function versionProbeFor(product, resolvedCandidate, resolvedCommand, repoRoot, pathValue) {
164
+ function versionProbeFor(product, resolvedCandidate, resolvedCommand, repoRoot, pathValue, skipNodeModules) {
122
165
  if (resolvedCandidate.versionSafe) {
123
166
  return { command: resolvedCommand, argv: ['--version'] };
124
167
  }
@@ -126,7 +169,10 @@ function versionProbeFor(product, resolvedCandidate, resolvedCommand, repoRoot,
126
169
  if (!candidate.versionSafe) {
127
170
  continue;
128
171
  }
129
- const found = locate(candidate.name, repoRoot, pathValue);
172
+ // The skip reaches here too, and it has to: a probe RUNS the binary, so
173
+ // resolving the version-safe sibling out of node_modules would execute a
174
+ // head-chosen program in order to answer a question about it.
175
+ const found = locate(candidate.name, repoRoot, pathValue, skipNodeModules);
130
176
  if (found !== null) {
131
177
  return { command: found.command, argv: ['--version'] };
132
178
  }
@@ -151,8 +197,9 @@ function baseName(file) {
151
197
  * named one specific file and falling back to something else would run a
152
198
  * different tool than the one they asked for.
153
199
  */
154
- export function resolveGateBinary(gate, repoRoot, pathValue) {
200
+ export function resolveGateBinary(gate, repoRoot, pathValue, options = {}) {
155
201
  const table = CANDIDATES[gate.product];
202
+ const skipNodeModules = options.skipNodeModules ?? false;
156
203
  if (gate.command !== undefined) {
157
204
  if (!existsAsFile(gate.command)) {
158
205
  throw new ResolveError(`gates.${gate.role}.command points at "${gate.command}", which is not a file. ` +
@@ -177,7 +224,7 @@ export function resolveGateBinary(gate, repoRoot, pathValue) {
177
224
  candidate: candidateName,
178
225
  versionProbe: candidate.versionSafe
179
226
  ? { command: process.execPath, argv: [gate.command, '--version'] }
180
- : versionProbeFor(gate.product, candidate, gate.command, repoRoot, pathValue),
227
+ : versionProbeFor(gate.product, candidate, gate.command, repoRoot, pathValue, skipNodeModules),
181
228
  };
182
229
  }
183
230
  return {
@@ -186,11 +233,11 @@ export function resolveGateBinary(gate, repoRoot, pathValue) {
186
233
  argvPrefix: prefix,
187
234
  source: 'policy',
188
235
  candidate: candidateName,
189
- versionProbe: versionProbeFor(gate.product, candidate, gate.command, repoRoot, pathValue),
236
+ versionProbe: versionProbeFor(gate.product, candidate, gate.command, repoRoot, pathValue, skipNodeModules),
190
237
  };
191
238
  }
192
239
  for (const candidate of table) {
193
- const found = locate(candidate.name, repoRoot, pathValue);
240
+ const found = locate(candidate.name, repoRoot, pathValue, skipNodeModules);
194
241
  if (found === null) {
195
242
  continue;
196
243
  }
@@ -200,7 +247,7 @@ export function resolveGateBinary(gate, repoRoot, pathValue) {
200
247
  argvPrefix: [...candidate.prefix],
201
248
  source: found.source,
202
249
  candidate: candidate.name,
203
- versionProbe: versionProbeFor(gate.product, candidate, found.command, repoRoot, pathValue),
250
+ versionProbe: versionProbeFor(gate.product, candidate, found.command, repoRoot, pathValue, skipNodeModules),
204
251
  };
205
252
  }
206
253
  return null;
package/dist/run.js CHANGED
@@ -1,7 +1,7 @@
1
1
  // One run across every enabled gate.
2
2
  import { compareFindings } from './envelope.js';
3
3
  import { EXIT_COULD_NOT_RUN, composeExitCode } from './exit-codes.js';
4
- import { preparationFailed, runGate } from './gate-runner.js';
4
+ import { gateProgramRefused, preparationFailed, refuseHeadControlledBinary, runGate, } from './gate-runner.js';
5
5
  import { resolveBaseRef } from './intent-base.js';
6
6
  import { frozenNativeContractPath, prepareIntent, } from './intent-prepare.js';
7
7
  import { GATE_ROLES, enabledGates, runsAtStage } from './policy.js';
@@ -23,10 +23,19 @@ function isPullRequestShaped(options, env) {
23
23
  }
24
24
  return resolveBaseRef({ ...(options.base === undefined ? {} : { base: options.base }), env }) !== null;
25
25
  }
26
- /** The intent-guard binary, or null when there is none to prepare with. */
26
+ /**
27
+ * The intent-guard binary, or null when there is none to prepare with.
28
+ *
29
+ * SAME SKIP AS runGate's, and it is not decoration: preparation SPAWNS this
30
+ * binary, before runGate has looked at anything. Resolving it out of
31
+ * node_modules on a pull-request run would execute a program the head chose,
32
+ * three times, before the program check the boundary rests on had run once.
33
+ */
27
34
  function intentBinary(gate, options) {
28
35
  try {
29
- return resolveGateBinary(gate, options.repoRoot, options.pathValue);
36
+ return resolveGateBinary(gate, options.repoRoot, options.pathValue, {
37
+ skipNodeModules: options.trustBase !== undefined,
38
+ });
30
39
  }
31
40
  catch (err) {
32
41
  // A ResolveError means the policy named a command that is not there.
@@ -174,6 +183,21 @@ export function runAll(policy, options) {
174
183
  // twice, once as a failed preparation and once as a missing gate,
175
184
  // would read as two problems where there is one.
176
185
  if (binary !== null) {
186
+ // BEFORE ANYTHING IS SPAWNED, and that ordering is the whole of it.
187
+ // runGate makes this check for every other gate and makes it first;
188
+ // for this one the preparation gets there earlier and runs the gate
189
+ // three times, so the check has to be made here as well or the
190
+ // boundary begins after the program the pull request chose has
191
+ // already run. Same builder as runGate's, so the enforce override
192
+ // cannot drift between the two places that produce this outcome.
193
+ const trustBaseRef = options.trustBase?.ref;
194
+ if (trustBaseRef !== undefined) {
195
+ const refusal = refuseHeadControlledBinary(options.repoRoot, trustBaseRef, binary);
196
+ if (refusal !== null) {
197
+ outcomes.push(gateProgramRefused(gate, trustBaseRef, refusal, binary));
198
+ continue;
199
+ }
200
+ }
177
201
  const prepared = prepareIntent({
178
202
  repoRoot: options.repoRoot,
179
203
  binary,
@@ -193,7 +217,15 @@ export function runAll(policy, options) {
193
217
  continue;
194
218
  }
195
219
  if (prepared.kind === 'failed') {
196
- outcomes.push(preparationFailed(gate, `the intent gate could not be prepared at the ${prepared.step} step: ${prepared.detail}`));
220
+ outcomes.push(preparationFailed(
221
+ // ENFORCED under a trust base, by the same argument as
222
+ // `refusedTrustBase` below: `enforce` is itself a control
223
+ // input, this gate produced no findings for it to be a
224
+ // decision about, and a pull-request run where the intent
225
+ // gate could not be prepared is one where nothing judged the
226
+ // intent. Reading the flag here let a base policy with
227
+ // enforce: false report the failure and still exit 0.
228
+ trustBaseRef === undefined ? gate : { ...gate, enforce: true }, `the intent gate could not be prepared at the ${prepared.step} step: ${prepared.detail}`));
197
229
  continue;
198
230
  }
199
231
  intent = prepared.preparation;
package/package.json CHANGED
@@ -1,8 +1,21 @@
1
1
  {
2
2
  "name": "@vaultcompass/conductor",
3
- "version": "0.3.0",
3
+ "version": "0.4.5",
4
4
  "type": "module",
5
5
  "description": "Umbrella installer and combined report over the independently installable guardrail gates.",
6
+ "keywords": [
7
+ "security",
8
+ "guardrails",
9
+ "sarif",
10
+ "code-scanning",
11
+ "github-actions",
12
+ "pre-commit",
13
+ "supply-chain",
14
+ "secrets-detection",
15
+ "cli"
16
+ ],
17
+ "homepage": "https://github.com/marketplace/actions/conductor-guardrail-gates",
18
+ "bugs": "https://github.com/vaultcompasshq/conductor/issues",
6
19
  "license": "MIT",
7
20
  "repository": {
8
21
  "type": "git",