GHSA-6F4G-XM8V-PGV6
Vulnerability from github – Published: 2026-11-10 15:56 – Updated: 2026-11-10 15:56Impact
The GitHub Actions runner treats certain text in a step's output as commands. The legacy form, ##[command], is recognized anywhere in a line. compose-lint's text report prints service names, environment keys and lines of the Compose file, and it does not neutralize ##[. Inside the compose-lint Action, a Compose file from a pull request can therefore:
- create fake error or warning annotations on the pull request (
##[error],##[warning]) - fold the rest of the log, including the real verdict, out of sight (
##[group]) - blank out chosen strings for the rest of the log (
##[add-mask])
GitHub has postponed removing the deprecated set-output log command, so the runner still honors it. A test run of the Action on a GitHub-hosted runner confirmed that a ::set-output line from a Compose file is processed. Whether it can override the Action's own sarif-written output, which decides whether the SARIF upload step runs, has not been tested.
The exit code and the pass/fail result are not affected. Azure Pipelines parses the same ##[ syntax, so the pip package is listed for anyone printing the text report in other CI systems.
Found during an internal pre-1.0 security review.
Patches
Fixed in 0.31.0. The Action stops the runner from processing workflow commands while compose-lint's output is printed, using a random stop-commands token. The text report also escapes ##[, and :: at the start of a line, in anything taken from the Compose file.
Workarounds
- In your own workflow, wrap the compose-lint step in
::stop-commands::<random token>and::<random token>::. - Use
--format jsonor--format sarifrather than the text report when linting untrusted pull requests.
Reproduction
services:
web:
image: "nginx:1.25"
environment:
"##[error]DB_PASSWORD": placeholder
"##[group]svc":
image: "nginx:1.25"
privileged: true
Run through the Action, this produces a fake error annotation and folds the remaining log lines, including the FAIL verdict, into a collapsed group.
{
"affected": [
{
"database_specific": {
"last_known_affected_version_range": "\u003c= 0.30.0"
},
"package": {
"ecosystem": "GitHub Actions",
"name": "tmatens/compose-lint"
},
"ranges": [
{
"events": [
{
"introduced": "0.2.0"
},
{
"fixed": "0.31.0"
}
],
"type": "ECOSYSTEM"
}
]
},
{
"database_specific": {
"last_known_affected_version_range": "\u003c= 0.30.0"
},
"package": {
"ecosystem": "PyPI",
"name": "compose-lint"
},
"ranges": [
{
"events": [
{
"introduced": "0"
},
{
"fixed": "0.31.0"
}
],
"type": "ECOSYSTEM"
}
]
}
],
"aliases": [
"CVE-2026-123369"
],
"database_specific": {
"cwe_ids": [
"CWE-117",
"CWE-150"
],
"github_reviewed": true,
"github_reviewed_at": "2026-11-10T15:56:59Z",
"nvd_published_at": "2026-11-09T18:17:25Z",
"severity": "MODERATE"
},
"details": "### Impact\n\nThe GitHub Actions runner treats certain text in a step\u0027s output as commands. The legacy form, `##[command]`, is recognized anywhere in a line. compose-lint\u0027s text report prints service names, environment keys and lines of the Compose file, and it does not neutralize `##[`. Inside the compose-lint Action, a Compose file from a pull request can therefore:\n\n- create fake error or warning annotations on the pull request (`##[error]`, `##[warning]`)\n- fold the rest of the log, including the real verdict, out of sight (`##[group]`)\n- blank out chosen strings for the rest of the log (`##[add-mask]`)\n\nGitHub has postponed removing the deprecated `set-output` log command, so the runner still honors it. A test run of the Action on a GitHub-hosted runner confirmed that a `::set-output` line from a Compose file is processed. Whether it can override the Action\u0027s own `sarif-written` output, which decides whether the SARIF upload step runs, has not been tested.\n\nThe exit code and the pass/fail result are not affected. Azure Pipelines parses the same `##[` syntax, so the pip package is listed for anyone printing the text report in other CI systems.\n\nFound during an internal pre-1.0 security review.\n\n### Patches\n\nFixed in 0.31.0. The Action stops the runner from processing workflow commands while compose-lint\u0027s output is printed, using a random `stop-commands` token. The text report also escapes `##[`, and `::` at the start of a line, in anything taken from the Compose file.\n\n### Workarounds\n\n- In your own workflow, wrap the compose-lint step in `::stop-commands::\u003crandom token\u003e` and `::\u003crandom token\u003e::`.\n- Use `--format json` or `--format sarif` rather than the text report when linting untrusted pull requests.\n\n### Reproduction\n\n```yaml\nservices:\n web:\n image: \"nginx:1.25\"\n environment:\n \"##[error]DB_PASSWORD\": placeholder\n \"##[group]svc\":\n image: \"nginx:1.25\"\n privileged: true\n```\n\nRun through the Action, this produces a fake error annotation and folds the remaining log lines, including the FAIL verdict, into a collapsed group.",
"id": "GHSA-6f4g-xm8v-pgv6",
"modified": "2026-11-10T15:56:59Z",
"published": "2026-11-10T15:56:59Z",
"references": [
{
"type": "WEB",
"url": "https://github.com/tmatens/compose-lint/security/advisories/GHSA-6f4g-xm8v-pgv6"
},
{
"type": "ADVISORY",
"url": "https://nvd.nist.gov/vuln/detail/CVE-2026-123369"
},
{
"type": "WEB",
"url": "https://github.com/tmatens/compose-lint/pull/918"
},
{
"type": "WEB",
"url": "https://github.com/tmatens/compose-lint/commit/672bd11145621f787c295fb11f3a859cdec1b38d"
},
{
"type": "PACKAGE",
"url": "https://github.com/tmatens/compose-lint"
},
{
"type": "WEB",
"url": "https://github.com/tmatens/compose-lint/releases/tag/v0.31.0"
}
],
"schema_version": "1.4.0",
"severity": [
{
"score": "CVSS:3.1/AV:N/AC:L/PR:N/UI:R/S:U/C:N/I:L/A:N",
"type": "CVSS_V3"
},
{
"score": "CVSS:4.0/AV:N/AC:L/AT:N/PR:N/UI:P/VC:N/VI:L/VA:N/SC:N/SI:L/SA:N/E:X/CR:X/IR:X/AR:X/MAV:X/MAC:X/MAT:X/MPR:X/MUI:X/MVC:X/MVI:X/MVA:X/MSC:X/MSI:X/MSA:X/S:X/AU:X/R:X/V:X/RE:X/U:X",
"type": "CVSS_V4"
}
],
"summary": "compose-lint: Compose file text can issue GitHub Actions workflow commands through the compose-lint Action"
}
Sightings
| Author | Source | Type | Date | Other |
|---|
Nomenclature
- Seen: The vulnerability was mentioned, discussed, or observed by the user.
- Confirmed: The vulnerability has been validated from an analyst's perspective.
- Published Proof of Concept: A public proof of concept is available for this vulnerability.
- Exploited: The vulnerability was observed as exploited by the user who reported the sighting.
- Patched: The vulnerability was observed as successfully patched by the user who reported the sighting.
- Not exploited: The vulnerability was not observed as exploited by the user who reported the sighting.
- Not confirmed: The user expressed doubt about the validity of the vulnerability.
- Not patched: The vulnerability was not observed as successfully patched by the user who reported the sighting.
The approach is described in our paper Mapping CVEs to MITRE ATT&CK Techniques: A Curated Gold-Set Classifier and the Limits of LLM-Assisted Label Expansion.
Browse all ATT&CK techniques and the vulnerabilities related to each.
Related by attack behaviour
Vulnerabilities whose description is nearest to this one in the vector space of the CIRCL/vulnerability-attack-technique-biencoder model. This is a similarity search over the bi-encoder space (plain cosine), not a classification, and it has no measured accuracy.