Offline evidence checker for CVE-2026-56850 — Node.js: HTTPS Agent PFX client identity collision. It compares Node.js runtime versions that an operator reports against the fixed releases named in the Node.js security advisory, and writes the result as CSV, JSON or script-free HTML.
The checker is a single Python 3.10+ file that uses only the standard library, so the executable can be copied and run on its own.
- Scope
- Requirements
- Quick start
- Synthetic report preview
- Usage
- Fixed versions
- Interpreting results
- Input formats
- Report formats
- Limits
- Exit codes and interruption
- Data handling
- Sources and provenance
- Documentation
- Contributing, security and license
This checker triages version evidence. It does not assess a deployed system.
- It is offline. It never contacts websites, executes Node.js or project code, or sends exploit payloads, and it does not scan the current machine.
- Its only evidence is the version (and optional build source and precondition observations) that you supply. A public URL is not an inventory.
- Hidden versions, custom or vendor builds, conflicting evidence, and release lines outside the rule snapshot remain
unknown. - Every result carries
exploitability: not_tested. No result establishes that the flaw is reachable or exploited. - A reported fixed version is not verification of a deployed patch. Version metadata is not a patch or build signature.
Use it only with inventories you are authorized to assess.
- Python 3.10 or later; no third-party packages.
- A Node.js version obtained from the deployed process or container (not from a workstation or lockfile), supplied directly, as a JSON inventory or as a CSV list.
CI runs the unit tests and --help on Python 3.10 and 3.14 on Ubuntu and Windows.
# Assess one runtime version and print JSON
python node_cve_2026_56850_check.py --version 22.23.2 --format json
# Assess the synthetic example inventory and save an HTML report
python node_cve_2026_56850_check.py --inventory examples/inventory.example.json --format html --output report.html
# Assess a CSV list and exit 1 if any result is a finding
python node_cve_2026_56850_check.py --csv-list inventory.csv --fail-on-findingRun python node_cve_2026_56850_check.py --help for the option list, or python -m unittest -v to run the local test suite.
Open the synthetic example report. Example data only; this is not a live assessment.
There is no automatic scan of the current machine. Supply an exact version obtained from the deployed process, then choose an input source and an output format.
You can combine any of these in one run; all results go into one report.
| Option | Input |
|---|---|
--version VERSION (alias --node-version) |
One Node.js version, such as 22.23.2; a leading v is accepted. Repeatable; each value describes a separate runtime. |
--inventory FILE |
JSON object or array of entries. Repeatable. See JSON inventory. |
--csv-list FILE |
CSV with a node_version column. Repeatable. See CSV list. |
At least one input is required. JSON, CSV and HTML reports are built from the same evidence.
| Option | Meaning |
|---|---|
--build-source SOURCE |
Provenance of the --version builds, at most 512 characters. unknown (the default), upstream, official and official_upstream are compared as upstream builds; any other value, such as a vendor name, yields vendor_build_unverified and a status of unknown. |
--shared-https-agent {true,false} |
Operator observation applied to --version inputs. |
--pfx-object-arrays {true,false} |
Operator observation applied to --version inputs. |
--different-client-certificates {true,false} |
Operator observation applied to --version inputs. |
--format {csv,json,html} |
Report format. The default is csv. |
--output FILE |
Save the report to a new file. The default is stdout. The parent directory must already exist. |
--force |
Replace an existing output file. Without it, an existing output path is an error. |
--brand NAME |
Organization name shown in the HTML report, at most 512 characters. The default is CompleteTech LLC. |
--fail-on-finding |
Exit 1 when any result is potentially_affected or unsupported_runtime. See Exit codes and interruption. |
--tool-version |
Print the checker name and version, then exit. (--version is the Node.js version, not the checker's.) |
The three observation options apply only to --version inputs; inventory and CSV entries carry their own observations. For example:
python node_cve_2026_56850_check.py --version 24.18.0 --build-source upstream --shared-https-agent true --format json- Reports are rendered fully in memory, then written to a temporary file in the output directory and published atomically.
- An existing output file is never replaced unless
--forceis explicit. --outputmust differ from every input file, even with--force. Output must be a regular file path; directories and dangling symbolic links are rejected.- Without
--force, publication uses a hard link so it cannot overwrite a file; on a filesystem without hard links the run fails with a message to choose another filesystem or use--force.
Reviewed 2026-10-24 against the official release advisory and linked fixed releases. Advisory-specific caveats are preserved below.
| Node.js major | Fixed version |
|---|---|
| 22.x | 22.23.2 |
| 24.x | 24.18.1 |
| 26.x | 26.5.1 |
The vendor does not specify an introduced version in the release advisory. The rules therefore treat earlier plain versions of each listed major as potentially_affected, and listed fix versions and later plain versions on the same major as fixed_version_reported.
For each listed major, versions below its fixed version require remediation. Unlisted, custom/backported, and prerelease builds need explicit evidence; do not label unknown builds safe.
How other inputs are classified:
- Other active or future majors remain
unknown(release_line_not_in_rules); a newer or unlisted major is not assumed to be patched. - End-of-life majors (0, 4–21, 23 and 25, per the Node.js release schedule as reviewed 2026-10-24; major 0 is an inference) produce
unsupported_runtimeas upgrade guidance. This is not a claim that this particular flaw exists in every historical build. - Vendor, distribution, backported or custom builds require vendor evidence and remain
unknown, because the upstream version does not show which patches the build contains. - Prereleases, nightly/rc builds and package suffixes remain
unknown(qualified_version_unverified).
Each input produces one result with a status and a more specific version_status. Both describe the reported version evidence, not the deployed system.
status |
Meaning |
|---|---|
potentially_affected |
The reported version is below the fixed version of its listed major and needs remediation review. |
fixed_version_reported |
The reported version is at or above the fixed version. This is not deployed-patch attestation; confirm that the running process or container actually uses that release. |
unsupported_runtime |
The major is end-of-life. This is upgrade guidance, not proof the flaw is present. |
unknown |
Evidence is missing or cannot be classified. Investigate; do not treat it as safe. |
Every result also has exploitability: not_tested, evidence_type: operator_reported_version, and an evidence_confidence of low (or insufficient when the status is unknown).
version_status values:
version_status |
Status | Cause |
|---|---|---|
below_fixed_version |
potentially_affected |
Below the fixed release of a listed major. |
at_or_above_fixed_version |
fixed_version_reported |
At or above the fixed release of a listed major. |
unsupported_release_line |
unsupported_runtime |
End-of-life major. |
release_line_not_in_rules |
unknown |
Major not covered by this dated rule snapshot. |
vendor_build_unverified |
unknown |
A vendor, distribution, custom or unrecognized build_source was reported. This check precedes the end-of-life check, so such builds are never reported as unsupported_runtime. |
qualified_version_unverified |
unknown |
Prerelease, nightly/rc, package-suffixed or other qualified version. |
unrecognized_or_partial_version |
unknown |
Not a plain MAJOR.MINOR.PATCH version. |
version_not_disclosed |
unknown |
No version was reported. Missing metadata does not show the runtime is unaffected. |
The advisory concerns shared HTTPS agents that are given PFX client identities. You may record three operator observations per input. Each is reported as reported_true, reported_false or not_reported.
| Observation | Meaning |
|---|---|
shared_https_agent |
shared https agent |
pfx_object_arrays |
pfx object arrays |
different_client_certificates |
different client certificates |
These are operator-supplied and unverified. They never change the version result and do not establish reachability.
Inputs are untrusted. Any invalid entry fails the whole run with exit 2 and no report is written.
One object or an array of objects. Each record requires integer schema_version: 1 and string node_version. Optional fields are runtime: "node", label, build_source, and the three observation keys above, each boolean or null. Unknown keys, duplicate keys, non-finite numbers and invalid types are rejected. Text fields are at most 512 characters. An entry without label is named <file name without extension>:<position> in the report; --version inputs are named Node.js <version>.
[
{
"schema_version": 1,
"label": "example-service",
"runtime": "node",
"node_version": "22.23.2",
"build_source": "unknown",
"shared_https_agent": null,
"pfx_object_arrays": null,
"different_client_certificates": null
}
]This is examples/inventory.example.json (synthetic).
A header row with a node_version column is required. Optional columns are label, build_source and the three observation columns; observations must be exactly true, false or blank. CSV lists have no schema_version or runtime column. Duplicate or undocumented columns, rows with the wrong number of fields, NUL characters, and malformed quoting (a quote inside an unquoted cell, text after a closing quote, or an unterminated quote) are rejected.
label,node_version,build_source,shared_https_agent,pfx_object_arrays,different_client_certificates
api-gateway,22.23.2,upstream,true,,
batch-worker,v24.18.0,,,,All formats share the same evidence and retain the canonical checker name, checker version, report schema (currently 1.0), rule date (rules_as_of) and a UTC generated_at. generated_at is the report generation time, not the time evidence was acquired. A single timestamp is reused for the whole run, including incomplete recovery.
- CSV (default). Every field is quoted, and values that could be read as spreadsheet formulas (starting with
=,+,-or@after leading whitespace, or with a tab or line break) are prefixed with an apostrophe.record_type=resultmarks completed assessments. When no result exists, a singlerecord_type=summaryrow preservesreport_completeand the rule metadata without an asset or verdict.tool,tool_versionandgenerated_atare appended to the other columns, including on empty summaries. NUL characters are rejected before a report is written. - JSON. An object with
tool,tool_version,report_schema_version,cve,rules_as_of,sources,complete,generated_atand aresultsarray. - HTML. A single script-free document with an embedded Content Security Policy, light and dark styles, status counts, results, the fixed-release table and the report schema.
--brandsets the organization name. An interrupted run is marked as such.
complete (JSON) and report_complete (CSV) are false when the run was interrupted and only the inputs assessed so far are included.
- Inputs must be regular UTF-8 files, at most 4 MiB each.
- At most 100 input files (
--inventoryand--csv-listcombined) and 10,000 entries in total, including--versionvalues. - At most 512 characters per text field, including
--version,--build-sourceand--brand. - Use stable local copies of inputs. File identity checks are not a sandbox against concurrent adversarial mutation.
| Exit code | Meaning |
|---|---|
0 |
The run completed and a report was written. Findings do not change this unless --fail-on-finding is used. |
1 |
--fail-on-finding was given and at least one result is potentially_affected or unsupported_runtime. unknown and fixed_version_reported do not trigger it. |
2 |
Usage, input, output or cleanup failure (an operational error, not an assessment result). This does not establish whether report bytes were already written; inspect the output before retrying. |
130 |
Interrupted. |
Interruption behavior:
- Before the output path is validated, nothing is written.
- Otherwise, completed results are written with
complete=falsewhen output is still possible. Only the inputs assessed so far appear, so assess the remaining inputs again. - If a complete report was already published, it is preserved and the run exits 130 without retrying.
- If file publication is indeterminate, the checker exits 130 without retrying. Inspect the output before retrying; the diagnostic does not establish whether publication succeeded.
- If the interrupt arrives while writing to stdout, the document may be truncated, and a second report is never appended.
Keep real inventories, targets, reports and secrets outside version control. Synthetic examples are in examples/ and docs/example-report.html.
Primary sources for the rule snapshot reviewed on 2026-10-24:
- https://nodejs.org/en/blog/vulnerability/july-2026-security-releases
- https://github.com/nodejs/security-wg/blob/main/vuln/core/187.json
- https://nodejs.org/en/blog/release/v22.23.2
- https://nodejs.org/en/blog/release/v24.18.1
- https://nodejs.org/en/blog/release/v26.5.1
End-of-life majors come from the Node.js release schedule, also reviewed 2026-10-24. The checker does not fetch these sources at runtime.
The embedded version engine is copied from node-cve-2026-105481-check commit de3161c4a8a7f7ca5c9e531df43b94e2805b52fd; no companion repository is required at runtime. Version metadata is not a patch or build signature.
- Contribution terms and security policy
- License (GNU AGPL-3.0) and dual licensing (AGPL-3.0 or a separate commercial license from CompleteTech LLC)
- Brand assets: CompleteTech LLC names, logos and the images in
docs/images/are not covered by the code license.
