Repository navigation
Releases
Codex Security GitHub Action 0.10.1
GitHub Action for Codex Security 0.10.1.
Installs @openai/codex-security@0.10.1.
uses: openai/codex-security@bf650710f36c16936324efe250a49cdd9f2da037 # action-v0.10.1Codex Security 0.10.1
Highlights
- Exclude paths from scans. Run
cs scan . --exclude-path examples/to exclude a directory from review in Standard, Deep, or diff scans. Repeat the flag for more paths, or setscan.exclude_pathsin your project configuration. Paths are literal and relative to the repository. The scanner can still read excluded files for context and filters out findings whose locations all fall under excluded paths. - Import scanner findings. Use
cs import continuumto collect AWS Continuum code issues with their supporting evidence. Attack-path analysis can use observations from the tested target when assessing reachability and severity. You can also import saved CrowdStrike Falcon Cloud Security IaC reports into your local workbench. - Deep Scan. When Codex blocks a worker request because of a misalignment policy violation, Deep Scan now stops and reports the refusal without retrying the request.
- Threat modeling. When a route appears in a test or on a deleted line, the threat-modeling instructions now ask the scanner to verify that the application still serves it.
Upgrade notes
If you used the bundled AWS Continuum helper collect_continuum.py, switch to cs import continuum ENVIRONMENT_ID --finding-id FINDING_ID --source-root /absolute/checkout --format json. For environments with multiple Git repositories or an S3 source, supply --repository-name with the AWS source name. Existing saved finding exports and patch captures remain compatible. Collect a finding with patches
All changes
What's Changed
Features
- feat(scanner-imports): import AWS Continuum code issues by @jameshiester-oai in #2619
- feat(imports): import CrowdStrike FCS IaC findings locally by @carlosmmatos-cs in #2589
- feat(examples): add synthetic security workflow benchmarks by @neharungta in #2618
- feat: add scan path exclusions by @ianw-oai in #2699
- feat(classify-severity): emit structured diagnostics on hosted records by @kmbroai in #2175
- feat(scanner-imports): retain validation evidence for individual Continuum findings by @mldangelo-oai in #2625
Fixes
- fix(plugin): preserve Windows helper exit codes by @JoaoMarcos44 in #2651
- fix(sdk): preserve symlink parent traversal in component paths by @mldangelo-oai in #2692
- fix(plugin): preserve Daybreak enrollment link attribution by @alandelong-oai in #2656
- docs(bulk-scan): show how --max-attempts works across two runs by @GautamSharma99 in #100
- fix(bulk-scan): reject receipt IDs missing from the campaign by @mldangelo-oai in #2677
- fix(deep-scan): stop on Codex misalignment policy refusals by @Shizoqua in #2410
- fix(history): avoid repeating findings in bulk scan matching by @GautamSharma99 in #149
- fix(findings): use saved comparison evidence across linked worktrees by @mldangelo-oai in #928
Documentation
- docs: clarify Deep Scan preflight sandbox limits by @dajiaohuang in #2611
- docs(plugin): keep repository links usable in standalone copies by @mldangelo-oai in #2664
- docs(providers): document Opper’s native profile requirements by @Felixkw12 in #1887
- docs(threat-model): check whether the app still serves a route by @Haiduongcable in #200
- docs: add a software citation for Codex Security by @mldangelo-oai in #2639
- docs(providers): document a tested xAI Responses gateway workaround by @mldangelo-oai in #2437
- docs(providers): configure OrcaRouter through a native file profile by @martinzudergaming-a11y in #736
- docs(examples): compare vulnerable and repaired invoice ownership checks by @neharungta in #2612
Other changes
- chore(deps): update the Python and Node devcontainer features by @dependabot[bot] in #2560
New Contributors
- @carlosmmatos-cs made their first contribution in #2589
- @neharungta made their first contribution in #2618
- @dajiaohuang made their first contribution in #2611
- @JoaoMarcos44 made their first contribution in #2651
- @Felixkw12 made their first contribution in #1887
- @martinzudergaming-a11y made their first contribution in #736
Full Changelog: npm-v0.10.0...npm-v0.10.1
Codex Security 0.10.0
Highlights
Codex Security 0.10.0 gives clearer guidance when a patch is blocked, lets interrupted Standard scans resume, and adds a budget for batch scans.
- Blocked patch results. The patch prompt asks whether any files changed, what stopped the fix, and what to do next. When a fix depends on work outside the checkout, the result explains why the external change is needed and where it belongs when known. #2609
- Standard scan setup. Standard scans start with the vulnerability audit and load host-specific instructions when needed. Focused investigators receive their own briefs, and scan history records the file count from the scan's inventory. #2046
- Resume interrupted Standard scans. Run
codex-security scans resume SCAN_IDto continue from a saved checkpoint in the original Codex session. The resumed scan retains its usage and budget. #2610 - Batch scan budgets. Set
--max-total-coston bulk or component scans to share one estimated budget across repositories or components, including retries and recoveries. In-flight work can exceed the estimate. #2592 - Scanner imports and publication. Import saved ArmorCode SAST findings into the local workbench. SDK callers can publish selected workbench finding versions to Cloud using retained evidence. #2608 · #2603
- Hosted remediation. A hosted patch or verification that ends at a different commit reports
revision_changedand retains its execution receipt. Retry from the requested revision. #2583
Upgrade notes
- POSIX paths through symlinks. Paths such as
link/../reponow follow filesystem traversal. To resume an older bulk campaign whose source was normalized differently, enter its exact saved source path in the CSV. Use a new output directory to scan the corrected physical target. #1709 - MCP registration scripts. Pass
--allif you relied on selecting all clients after automatic detection found none, or pass--agent NAMEfor one client. Use--package NAMEfor bare npm package names previously passed to--command. #2591 - Archived scanner captures and Cloud retries. Recollect Semgrep Platform v1 and Snyk REST v1 captures with the current collectors. Finish interrupted Cloud submissions with the client that created them. If the outcome is uncertain, check the inbox job or repository records before submitting again. #2602
- Patch follow-ups with scan budgets. Run patch preparation separately when a scan has a cost limit. The scan rejects an explicit patch follow-up and skips its interactive patch offer. #2592
All changes
What's Changed
Breaking changes
- fix(paths)!: preserve physical parent traversal by @mldangelo-oai in #1709
- fix(cli)!: make MCP targets and launchers explicit by @mldangelo-oai in #2591
- refactor(imports)!: retire historical reader formats and Cloud checkpoints by @mldangelo-oai in #2602
Features
- feat(cli): expose identifiers for saved scan follow-up by @mldangelo-oai in #2577
- feat(cli): make project setup and authentication easier to inspect by @mldangelo-oai in #2581
- feat(imports): publish versioned workbench findings through the shared Cloud transport by @mldangelo-oai in #2603
- feat(plugin): guide Codex through scanner imports and workbench publication by @mldangelo-oai in #2604
- feat(scanner-imports): add ArmorCode findings support by @mldangelo-oai in #2608
- feat(import): include AWS-generated patches in Continuum findings by @mldangelo-oai in #2615
- feat(scans): resume interrupted Standard scans by @ianw-oai in #2610
Fixes
- fix(deep-scan): resolve home-relative project config paths on POSIX by @mldangelo-oai in #2505
- fix(windows): preserve literal paths when loading scan policy by @mldangelo-oai in #2510
- fix(plugin): select a SQLite-capable Node runtime on Linux by @mldangelo-oai in #2499
- fix(package): prevent stale SDK archives on long Windows paths by @mldangelo-oai in #2511
- fix(sdk): follow directory links in selected project config paths by @mldangelo-oai in #2514
- fix(windows): report filesystem module failures during output validation by @mldangelo-oai in #2516
- fix(package): build the bundled plugin in long Windows checkouts by @mldangelo-oai in #2508
- fix(cli): preserve Linux install path case in update notices by @mldangelo-oai in #2521
- fix(windows): resume Deep Scans interrupted before their first artifact by @mldangelo-oai in #2515
- fix(sdk): create scan output for long POSIX repository names by @mldangelo-oai in #2527
- fix(windows): retry rolled-back plugin cache activation by @mldangelo-oai in #2528
- fix(plugin): save artifacts with long POSIX filenames by @mldangelo-oai in #2533
- fix(package): explain the Alpine tar prerequisite in release checks by @mldangelo-oai in #2536
- fix(scans): archive long POSIX output directory names by @mldangelo-oai in #2529
- fix(workbench): identify replaced checkout parent directories by @mldangelo-oai in #2537
- fix(workbench): register nested Windows scans with long Git metadata paths by @mldangelo-oai in #2534
- fix(plugin): preserve closed helper stdin under BusyBox ash by @mldangelo-oai in #2540
- fix: preserve Continuum import retries and active sign-in credentials by @mldangelo-oai in #2531
- fix(runtime): find system Python for Linux accounts without a home by @mldangelo-oai in #2509
- fix(release): allow more time for npm package availability by @mldangelo-oai in #2532
- fix(cli): use the singular for one review round in Deep Scan stop reasons by @Shizoqua in #2443
- fix(knowledge-base): skip DOCX content control placeholder text by @Shizoqua in #2490
- fix(plugin): publish new patch files on musl without hard links by @mldangelo-oai in #2530
- fix(workbench): preserve Windows checkout identity across Python versions by @mldangelo-oai in #2541
- fix(sdk): register local marketplaces in Codex homes containing a hash by @mldangelo-oai in #2547
- fix(sdk): preserve whitespace in POSIX state directory overrides by @mldangelo-oai in #2548
- fix(container): record the Wolfi Canvas 1.1.0 source and Skia notice by @mldangelo-oai in #2549
- fix(plugin): preserve trailing whitespace in POSIX scan roots by @mldangelo-oai in #2552
- fix(runtime): honor literal POSIX executable override paths by @mldangelo-oai in #2551
- fix(workbench): preserve explicit usage database paths by @mldangelo-oai in #2556
- fix(windows): grant scan workers explicit source read access by @mldangelo-oai in #2545
- fix(scan): verify assumed-unchanged files for committed diffs by @mldangelo-oai in #2554
- fix(cli): resolve temporary paths before patching a saved checkout by @mldangelo-oai in #2558
- fix(sdk): stage knowledge in the selected temporary directory by @mldangelo-oai in #2559
- fix(scan): isolate Git pathspec modes for file inventories by @mldangelo-oai in #2565
- fix(native): read sealed artifacts in searchable Linux directories by @mldangelo-oai in #2564
- fix(workbench): pin Git metadata on another Windows fixed d...
Codex Security GitHub Action 0.10.0
GitHub Action for Codex Security 0.10.0.
Installs @openai/codex-security@0.10.0.
uses: openai/codex-security@3bde66eda1a5c6b0c4cd245f51c16424c9374afb # action-v0.10.0Codex Security GitHub Action 0.9.1
GitHub Action for Codex Security 0.9.1.
Installs @openai/codex-security@0.9.1.
uses: openai/codex-security@a98b35d7eef9f85c298c86229445bb64ac427aaf # action-v0.9.1Codex Security 0.9.1
Highlights
- The npm package includes Developer ID-signed native addons for Apple Silicon and Intel Macs. This lets Codex Desktop load the addon when preparing review items.
- Deep Scan runs independent Standard reviews in batches. Workers can share leads and verify each other's evidence. The final report groups observations that have the same fix and retains the source locations from each review. Deep Scan guide
- External scanner imports are enabled by default in the CLI, SDK, and plugin. You can preview supported reports, save findings locally, and publish supported findings to Cloud. The scanner guides list the accepted formats and which inputs are local-only. Scanner import guides
scans logsfinds sessions faster in large compressed histories. The GitHub Action also includes scan scope limitations in its job summary, including when a scan returns zero findings.
Upgrade notes
- Deep Scan budgets.
--stop-after-no-newnow counts consecutive completed Standard reviews that add no findings. If you depend on a particular run length, review this threshold alongside--max-discovery-runsand--max-time. Deep Scan settings - Scanner imports. Remove
CODEX_SECURITY_ENABLE_EXTERNAL_IMPORTSfrom launch scripts. A legacy value of0no longer disables imports. Scanner import guides - Linux Compose deployments. The bundled seccomp profile restricts direct socket families and blocks
ptrace,process_vm_readv, andprocess_vm_writev. If a custom tool needs one of these operations, test it against your deployment's profile before upgrading. Container guide - Source builds. Sparse SDK checkouts must include
plugins/codex-security/mcp-app/src/deep-scan-settings.ts. Upgrade or roll back the bundled plugin as a unit. Source build change - Package verification. The source package checker rejects sparse tar entries. If an unsigned candidate was built with a sparse-aware archiver, recreate it with
pnpm packfromsdk/typescript. Use the original published archive for release recovery. Package verification change
All changes
What's Changed
Breaking changes
- feat!: deep scan upgrade by @ianw-oai in #2118
- refactor(package)!: reject sparse entries in release tarballs by @mldangelo-oai in #2456
- refactor!: consolidate scan, workbench and release rules by @mldangelo-oai in #2452
Features
- feat: expose durable scan lifecycle telemetry by @mldangelo-oai in #2274
- feat(imports): import saved Semgrep selections through the CLI by @mldangelo-oai in #2166
- feat(imports): collect Semgrep findings through Guardian and the Web API by @mldangelo-oai in #2167
- feat(imports): collect Snyk Code and Open Source findings locally by @mldangelo-oai in #2168
- feat(imports): publish hosted scanner findings to Cloud by @mldangelo-oai in #2085
- feat(container)!: restrict sockets and process tracing in Compose by @MauricioPerera in #160
- feat(import): collect and import Checkmarx SAST findings by @mldangelo-oai in #2269
- feat(imports): read native Trivy dependency reports in the SDK by @mldangelo-oai in #2087
- feat(deep-scan): share investigation notes between workers by @mldangelo-oai in #2265
- feat(imports): import Trivy findings locally and to Cloud by @mldangelo-oai in #2276
- feat(imports): import Tenable One code repository findings locally by @mldangelo-oai in #2343
- feat(imports): enable external scanner imports by default by @mldangelo-oai in #2360
- feat(imports): publish Guardian findings and clarify scanner evidence by @mldangelo-oai in #2377
- test(scans): cover private endpoint profile cleanup and replay by @TomCN0803 in #125
Fixes
- fix(action): use the singular for one additional finding in the job summary by @WenJing95 in #2252
- fix(dashboard): keep scrolled history in place at the history cap by @LittleYier in #2266
- fix(knowledge-base): skip macOS metadata files in knowledge-base folders by @Shizoqua in #2270
- fix(knowledge-base): keep non-breaking hyphens in DOCX text by @Shizoqua in #2272
- fix(imports): honor bounded Retry-After on hosted scanner reads by @mldangelo-oai in #2086
- fix(runtime): handle disconnected child stdin on macOS by @mldangelo-oai in #2275
- fix(deep-scan): stop workers when the plan does not include Codex by @Shizoqua in #1843
- fix(knowledge-base): skip LibreOffice lock files in knowledge-base folders by @Shizoqua in #2291
- fix(sdk): pass severity record efforts through to the host by @Shizoqua in #1842
- fix(knowledge-base): ignore DOCX tab stop definitions when extracting text by @Shizoqua in #2284
- fix(deep-scan): recognize usage limits reported during compaction by @Shizoqua in #1844
- fix(knowledge-base): omit DOCX drawing positions and sizes from extracted text by @Shizoqua in #2051
- fix(targets): report a failed ignored-but-tracked file listing by @kevin9327 in #864
- fix(targets): check resolved path types before scan registration by @sylvesterkaczmarek in #549
- fix(cost): preserve worker activity when a session log is unreadable by @feiiiiii5 in #970
- fix(activity): preserve shell-escaped whitespace in repository paths by @RakamakadoN in #2163
- fix(runtime): protect HF_TOKEN and defer the Hugging Face scan preset by @NielsRogge in #277
- fix(findings): align severity badges in finding tables by @mldangelo in #122
- fix(hooks): use trusted Git when installing hooks by @GautamSharma99 in #92
- fix(deep-scan): preserve distinct findings and bound duplicate comparisons by @mldangelo-oai in #2313
- fix(bulk-scan): reject invalid saved attempt numbers by @mldangelo-oai in #2292
- fix(cli): explain the cost limit in scan dry runs by @mldangelo-oai in #2295
- fix(windows): inspect long credential-home descendant paths by @mldangelo-oai in #2294
- fix(knowledge-base): preserve documents with AppleDouble-style names by @mldangelo-oai in #2312
- fix(sdk): persist canceled scans with their measured cost by @mldangelo-oai in #2307
- fix(imports): preserve Guardian repository identity during replay by @mldangelo-oai in #2308
- fix(deep-scan): recover the latest public findings on interruption by @mldangelo-oai in #2310
- fix(sdk): forward component scan lifecycle and deep progress by @mldangelo-oai in #2317
- fix(deep-scan): align worker artifact validation with publication by @mldangelo-oai in #2315
- fix(imports): bound remote error bodies and retain HTTP failure context by @mldangelo-oai in #2320
- fix(deep-scan): checkpoint accepted reviews before honoring stop conditions by @mldangelo-oai in #2318
- fix(dashboard): keep the viewport anchored during worker updates by @mldangelo-oai in #2298
- fix(deep-scan): attribute comparison sessions to usage and logs by @mldangelo-oai in #2299
- fix(deep-scan): require consistent evidence before merging reports by @mldangelo-oai in https://github.com/open...
Codex Security 0.9.0 (source only)
Source-only release. The source archives below contain the original 0.9.0 release commit. The 0.9.0 build stopped during macOS signing, before npm publication. For the signed CLI and SDK package, use Codex Security 0.9.1.
Highlights
- Deep Scan runs independent Standard reviews in batches. Workers can share leads and verify each other's evidence. The final report groups observations that have the same fix and retains the source locations from each review. Deep Scan guide
- External scanner imports are enabled by default in the CLI, SDK, and plugin. You can preview supported reports, save findings locally, and publish supported findings to Cloud. The scanner guides list the accepted formats and which inputs are local-only. Scanner import guides
scans logsfinds sessions faster in large compressed histories. The GitHub Action also includes scan scope limitations in its job summary, including when a scan returns zero findings.
Upgrade notes
- Deep Scan budgets.
--stop-after-no-newnow counts consecutive completed Standard reviews that add no findings. If you depend on a particular run length, review this threshold alongside--max-discovery-runsand--max-time. Deep Scan settings - Scanner imports. Remove
CODEX_SECURITY_ENABLE_EXTERNAL_IMPORTSfrom launch scripts. A legacy value of0no longer disables imports. Scanner import guides - Linux Compose deployments. The bundled seccomp profile restricts direct socket families and blocks
ptrace,process_vm_readv, andprocess_vm_writev. If a custom tool needs one of these operations, test it against your deployment's profile before upgrading. Container guide - Source builds. Sparse SDK checkouts must include
plugins/codex-security/mcp-app/src/deep-scan-settings.ts. Upgrade or roll back the bundled plugin as a unit. Source build change - Package verification. The source package checker rejects sparse tar entries. If an unsigned candidate was built with a sparse-aware archiver, recreate it with
pnpm packfromsdk/typescript. Use the original published archive for release recovery. Package verification change
Codex Security 0.8.3
Highlights
- update SDK and MCP parser dependencies (#2141)
- do not split surrogate pairs in truncated report text (#2133)
- update vulnerable triage dependencies (#2142)
- defer CLI and plugin validators until first use (#2140)
- explain package provenance (#361)
- speed up plugin setup and defer CLI prompts (#2146)
- track Black Duck SCA project-version vulnerabilities (#2098)
- avoid claiming output was kept after scan failures (#359)
- track Veracode static application policy flaws (#2099)
- accept UTF-8 BOM in JSON project files (#980)
- avoid duplicate plugin setup inspection and defer CLI YAML (#2149)
- accept UTF-8 BOMs in scan inputs and calibration datasets (#2151)
- preserve IPv6 verification URL brackets (#232)
- bump the third-party group across 2 directories with 2 updates (#2157)
- bump ruff from 0.17.1 to 0.17.2 in /plugins/codex-security (#2153)
- bump docker/login-action from 4.6.0 to 4.7.0 (#2155)
- bump the codex group across 3 directories with 2 updates (#2154)
- keep whole characters when shortening publish progress (#2158)
- preserve local scan startup diagnostics (#2115)
- name the outermost rejected Windows ancestor (#2065)
- count only CR and LF as assessment line breaks (#2147)
- retain the reviewed surface count for completed Deep scans (#2161)
- show highest-severity findings first (#2160)
- bump taiki-e/install-action from 2.87.30 to 2.87.31 (#2156)
- align Unicode repository columns (#2148)
- link Markdown evidence to plain source lines (#2032)
- update MCP SDK to 1.32.2 and Stryker to 10.1.1 (#2170)
- import SonarQube Cloud security findings (#2162)
- share saved coverage logic and checkpoint fixtures (#2185)
- organize external scanner integration guides (#2189)
- index SARIF artifact URIs once per run (#2191)
- normalize GitHub scanner evidence for its destination (#2192)
- reuse Wiz replay projections for package reports (#2193)
- share local Git metadata discovery (#2194)
- require exhaustive scanner reader dispatch (#2195)
- normalize selected Semgrep Code findings for local import (#2164)
- share optional field parsing across GitHub and Wiz (#2198)
- share scanner Cloud publication and recovery guidance (#2199)
- share SARIF rule indexing and reference checks (#2200)
- isolate rejected publication progress callbacks (#2203)
- share connection and error handling for finding imports (#2204)
- avoid duplicate Git metadata discovery (#2205)
- skip Cloud number formatting for local evidence (#2202)
- give each scanner import a focused guide (#2201)
- reuse native record checks for Black Duck components (#2207)
- reduce Sonar issue collection requests (#2208)
- show the SDK fallback for finding imports (#2209)
- show how to retrieve selected scanner evidence (#2210)
- share SARIF result context for Semgrep and Endor (#2219)
- share the repository scope for external finding reads (#2221)
- choose configured tools for scanner and Cloud findings (#2213)
- parse fenced triage results in one pass (#2214)
- simplify publication progress and scan selection (#2218)
- share the latest remediation attempt query (#2215)
- share patch and verify-fix input selection (#2217)
- share smoke inputs and native artifact upload settings (#2220)
- share artifact path validation during finalization (#2222)
- remove private Wiz finding type re-exports (#2229)
- remove unreachable component dashboard status cases (#2230)
- reuse validated patch risk summaries in draft reviews (#2228)
- remove unused permissive Git-tree lookup (#2227)
- remove retired GitHub publication commands (#2226)
- share SARIF rule reference validation (#2235)
- simplify provider import examples (#2234)
- skip Cloud replay projections for local Wiz findings (#2236)
- accept imported evidence in records deduplication (#2241)
- use the singular for one finding or reviewed file (#2239)
- preserve PDF line breaks during text extraction (#2196)
- skip index paths under directories replaced by files (#2182)
- use the singular for one repository in bulk scan prompts (#2174)
- spell numeric ranges with en dashes, not tildes (#2095)
- keep every scroll step in the finding details pane (#2016)
- align the component cost column with its header (#2004)
- bump taiki-e/install-action from 2.87.31 to 2.87.32 (#2244)
- distinguish library defaults from caller obligations (#2245)
- preserve boxed numeric evidence across retries (#2224)
- restore concrete discovery guidance for Standard and Deep scans (#2246)
- reuse repository membership checks across scans (#2249)
- preserve permission errors from plugin installation (#2247)
- share terminal grapheme handling (#2232)
- prepare Veracode selection filters once per report (#2248)
- index saved sessions by parent when reading scan logs (#2237)
- share SARIF evidence projection for Semgrep and Endor (#2233)
- share projection and validation logic (#2183)
- reconcile unacknowledged findings uploads (#2211)
- reuse scoped GitHub connections for finding intake (#2223)
- clarify the post-scan findings review prompt (#2255)
- remove Daybreak latest alias pricing (#2165)
- opt in to the remote rubric for standalone reclassification ([#2257](https://gith...
Codex Security GitHub Action 0.8.3
GitHub Action for Codex Security 0.8.3.
Installs @openai/codex-security@0.8.3.
uses: openai/codex-security@90c85a3617e06eb5b4ac139778ef06581f44e532 # action-v0.8.3Codex Security 0.8.2
Highlights
Codex Security 0.8.2 adds threat-model explorers and Amazon Inspector imports, and improves scan progress and finding review.
- Threat-model explorers. Turn an existing threat model into a shareable HTML explorer with interactive architecture, scenarios, controls, and source evidence. Export HTML or SVG from the bundled skill.
- Amazon Inspector imports. With external imports enabled, import a complete Inspector Code Security repository findings export while retaining each finding's identity across imports. See the Inspector import guide.
- Dedupe corrections. Hosted reviewers can correct rejected decisions in the original open review when the host supports pending-call validation.
- Interactive scans. Finding review shows critical findings first, followed by high, medium, and low severity. Component tables show saved review progress, and scan titles resolve relative directory names.
- Output directories. Existing scan output directories keep their permissions across scans, saved-result reads, and resume. Newly created directories remain private by default.
- Container releases. The release workflow builds Debian and Wolfi images from the same verified scanner package and frozen runtime dependency lock. See the container guide.
Upgrade notes
No migration is required for existing CLI workflows. Hosted integrations that opt into pending-call validation must keep the original review session open until they handle the CLI's feedback; existing hosts can continue returning final results.
Inspector imports remain opt-in. Cloud publication requires a connected GitHub repository and a deployment that accepts the Inspector provider. The import guide describes repository matching and complete-export requirements.
All changes
What's Changed
Features
- feat: resolve dependency alerts by @mldangelo-oai in #2084
- feat(import): track Amazon Inspector repository vulnerabilities by @mldangelo-oai in #2097
- feat: add a threat-model explorer by @fwph-oai in #2112
- feat(container): release Debian and Wolfi from the verified scanner package by @mldangelo-oai in #2116
Fixes
- fix: allow existing scan output directory permissions by @suhailpatel-oai in #2120
- fix(dashboard): show saved review progress in the component table by @WenJing95 in #2111
- fix(action): check RUNNER_TEMP before resolving it by @LittleYier in #2125
- fix(auth): keep login URLs when OSC sequences end with ST by @LittleYier in #2124
- fix: ship explorer references as standalone vector images by @mldangelo-oai in #2128
- fix: resolve directory names in scan titles by @suhailpatel-oai in #2121
- fix: sort interactive findings by severity by @suhailpatel-oai in #2135
- fix(dedupe): correct rejected decisions in the open records review by @kmbroai in #2139
Documentation
- docs: explain how to verify a committed diff scan inventory by @mldangelo-oai in #2122
- docs: add published docs to the release checklist by @mldangelo-oai in #2094
Other changes
- ci: bump taiki-e/install-action from 2.87.29 to 2.87.30 by @dependabot[bot] in #2108
- ci: bump actions/setup-python from 7.0.0 to 7.1.0 by @dependabot[bot] in #2106
- chore(deps): bump @hono/node-server from 2.1.1 to 2.1.5 in /plugins/codex-security/mcp-app by @dependabot[bot] in #2123
- chore(native): document exact dependency pins by @mldangelo-oai in #2126
- ci: bump actions/attest from 4.2.1 to 4.2.2 by @dependabot[bot] in #2131
- chore(deps): bump @linear/sdk from 98.0.0 to 99.0.0 in /sdk/typescript by @dependabot[bot] in #2132
- chore(deps): bump napi from 3.14.2 to 3.15.0 in /plugins/codex-security/native by @dependabot[bot] in #2130
- chore(deps): bump napi-derive from 3.6.12 to 3.6.13 in /plugins/codex-security/native by @dependabot[bot] in #2129
- refactor: simplify scanner imports and SDK test setup by @mldangelo-oai in #2136
- refactor: simplify workbench and Deep Scan processing by @mldangelo-oai in #2137
New Contributors
- @suhailpatel-oai made their first contribution in #2120
- @fwph-oai made their first contribution in #2112
Full Changelog: npm-v0.8.1...npm-v0.8.2