* feat(cli): add independently reviewed dependency checks
* Compare frozen Python advisory versions as releases
* fix(cli): bind unnamed root manifests to observed services
* test(cli): evaluate dependency investigations across public fixtures
* test(sdk): bound the budget wait after the mock scan starts
* test(cli): exercise MCP evidence and missing caller access
* Bind graded citations to the approved use evidence
* test: restore dependency investigation quality controls
* fix(cli): restore dependency investigation and report compatibility
* test: grade malformed dependency baseline answers
* fix(cli): scope headless tool access to the investigator
* test(cli): lock down reviewer and external evidence tool access
* fix(cli): keep linked evidence tied to its originating call
* Reconcile application mappings and expose original evidence hashes
* test: verify selected catalog rows against original replies
* fix(cli): recover incomplete reviews and preserve selected evidence rows
* fix(cli): preserve wire evidence and verify finding citations
* Finalize from approved submissions and explain remaining uses
* fix(cli): restore coordinator recovery, independent review and report contracts
* Review withdrawals under the original dependency assessment ID
* test(cli): compare trace progress with completed search scopes
* Recheck dependency coverage after concurrent investigation changes
* Preserve exploratory scope explanations during dependency review
* feat(cli): track shared source work in dependency investigations
* Include shared source work in final coverage checks
* Require deciding evidence on the caller it establishes
* Preserve generic MCP search outcomes and scopes in dependency reports
* Share pinned source notes between dependency investigators
* Ship the source worklist and search audit modules
* refactor(cli): name the dependency coordinator for its ability to resume
* refactor(cli): simplify the tools for dependency investigations
* fix(cli): keep investigation tool arguments visible to Codex and reuse the evidence store
* fix(cli): send missing verdict citations back to the reviewer
* fix(cli): verify caller bindings and input flow across investigation steps
* fix(cli): require direct citations for asserted runtime facts on every use
* test(cli): verify failed catalog lookups without a timed source scan
---------
Co-authored-by: utsav-oai <348536218+utsav-oai@users.noreply.github.com>
* test(plugin): apply saved patch bytes on Windows
* test(ci): provide ripgrep for the Windows MCP smoke test
* test(sdk): preserve env utility name in inventory fixture
(cherry picked from commit 5ce7e7c53c)
* test(plugin): make Deep Scan worker fixtures portable
* test(ci): exercise Deep Scan worker fixtures on Windows
* test(plugin): isolate worker PATH fixtures from the SDK
* test(plugin): format the worker search path fixture
* ci(release): share npm setup and verification
Use the same local composite action after checkout in both release jobs. Keep the Node and registry settings, Socket Firewall installation and next-step PATH verification in one place. Each job retains its own checkout, credentials and publication controls. Existing tags continue to use the workflow in their release commit. No public interface migration is required.
Both release suites pass (318 tests), including 312 tests with the separate bootstrap and release-history changes applied. Five mocked setup runs checked success and failures during installation, version validation and executable selection. Compared workflow selection for npm-v0.8.2 and npm-v0.8.3 and linted both workflows. These checks did not dispatch a release or publish a package.
* ci(release): scope the workflow lint exception to the verified npm path
Moving the shared setup into a composite action brings its existing PATH write under the github-env audit. Both release jobs install the pinned npm from the firewall, verify the executable in a fixed directory under RUNNER_TEMP and recheck it on PATH in the next step. Explain that input on the installation step.
Reproduced the initial diagnostic with zizmor 1.30.1. Its repository-wide workflow check now passes, and it still reports an unchecked PATH write added to a neighboring fixture step. Both release automation and release PR suites pass (429 tests).
Read published versions, release tags, and reachable tags in the publisher before generating notes. Keep the previous tag and latest decision in their native types and remove the intermediate workflow outputs. The internal release-history command had one workflow caller. No public API or client migration is required.
Verified with 318 release automation and release progress workflow tests. Thirteen synthetic comparisons of the previous and revised pipelines matched command arguments, generated notes, and exit codes for initial releases, backfills, partial histories, and failures.
The public package has already shipped 0.1.0 and npm permanently reserves
that name and version. Remove the first-publish token step and the local
adapter that treated an E404 as an empty release history. Both workflows
stop on lookup failure. Normal publication still uses trusted publishing.
Historical GitHub release verification keeps its original recovery path.
Across 22 synthetic release-cut and publisher comparisons, only the two
initial-404 cases changed. The candidate passed 407 release, Action, and
container tests. The historical 0.1.0 npm and GitHub archives matched
byte for byte. The pinned npm signature audit and both revisions'
provenance and publication checks accepted the same archive and run.
No package user or downstream release command needs to change.
The protected workflow schedules one main baseline or a manually selected
PR revision. Carry that SHA and optional PR number through job outputs,
and recheck the same identity after approval. The trusted cache gates,
visible job and artifact names, and revision-specific concurrency persist.
Check actual PR, baseline and skipped CLI outputs. The 5 resolver and 10
workflow tests passed on both revisions. Also compared both resolvers
across 504 cases and the emitted CLI outputs and workflow expressions
across 33 dispatch, eligibility, approval and cache scenarios.
Compute job selection directly for changes limited to existing Python tests. Keep their full Ubuntu version matrix, Windows Python job, package checks and workflow checks. Each required-job output now has one definition.
Compared every output and matrix across 68,832 path and event combinations, including the 1,515 tracked paths and mixed changes. All 27 CI selection and source verification fixtures passed on the parent and the change.
* ci(deps): update Bun to 1.4.4
* test(ci): align the Bun workflow version assertion
* chore(deps): align remaining Bun toolchain pins
* fix(ci): preserve test timeouts with Bun 1.4 options
---------
Co-authored-by: Michael D'Angelo <mdangelo@openai.com>
* fix: finish Windows native wrapper publications before reporting failure
* test(plugin): verify Windows wrapper retries and cleanup before rejection
---------
Co-authored-by: mldangelo-oai <269034524+mldangelo-oai@users.noreply.github.com>
* fix: give bundled MCP helpers a package scope on Windows UNC shares
* ci: exercise the bundled UNC helper on Windows Node 26
* ci: keep the Windows machine-policy check on Node 22
---------
Co-authored-by: mldangelo-oai <269034524+mldangelo-oai@users.noreply.github.com>
* test(deep-scan): make worker lifecycle fixtures deterministic on Windows
* ci: cover Deep Scan cleanup and deadlines on Windows
---------
Co-authored-by: mldangelo-oai <269034524+mldangelo-oai@users.noreply.github.com>
* test(mcp): resolve the native binding in candidate fixtures
* ci: exercise candidate normalization on the Windows runner
---------
Co-authored-by: mldangelo-oai <269034524+mldangelo-oai@users.noreply.github.com>
Use the shared Node reader to verify sealed scan artifacts before tracking selects findings. Direct callers of the former Python helper can use the plugin launcher.