Finding and attribution
In gh-aw versions before v0.91.12, MCP handlers appended untrusted client strings into child-process argv without a `--` separator or leading-hyphen rejection. A lower-trust MCP client could inject supported flags to bypass workflow security controls and perform credential-backed GitHub operations. GHSL assigns identifier GHSL-2026-180 and the repository advisory GHSA-f7r8-5pqp-q5p4; no CVE is assigned.
Product: GitHub gh-aw CLI and MCP server. AI system: GitHub Security Lab Taskflow Agent. Team / organization: GitHub Security Lab and GitHub.
CVE identifiers: Not established in the cited evidence.
AI and human contribution (source account): The published repository advisory explicitly says Taskflow Agent discovered the issue and GHSL manually verified it. It identifies the AI system and human verification separately; no priority claim is needed.
Human involvement is only described where the source account establishes it; unknown involvement does not mean none. Vendor confirmation of a flaw does not confirm every claim about AI contribution.
Discovery context: The advisory says the fix was included in v0.91.12 published October 14, 2026; that version is marked pre-release.
Remediation status: fix documented. Patched in gh-aw v0.91.12, published October 14, 2026 and marked pre-release. The advisory recommends v0.91.12 or later; downstream deployment is not verified.
A source-documented fix does not establish installed patches or current safety. Source checking does not independently reproduce the finding or establish observed exploitation.
Disclosed 14 Oct 2026. AI attribution dated 14 Oct 2026. Sources checked 14 Nov 2026; these are separate dates, not a live installation check.
Recorded CVSS: Not available · no assessment established here. Documented technical effects: data modification, other.
Severity, consequences and exploitation evidence
Case grouping uses the highest recorded CVSS score, with its version-specific severity band. It is descriptive, not a combined or editorial risk index. All recorded assessments remain below; scores from different versions and sources may disagree. CVSS specifications define version-specific scoring and severity bands.
No source-backed CVSS assessment is established in this catalog.
Case-level evidence: data modification, other.
Prerequisites / affected scope: A less-trusted MCP client able to reach the local server can inject argv flags that bypass intended restrictions, perform GitHub writes (push branches/open PRs), disable scanners, alter workflows/lock files, and write outside `.github/workflows`. The advisory also says a specific `mcp-inspect` argument can launch the MCP Inspector via npx; it does not establish arbitrary attacker-chosen code execution..
Exploitation evidence: The advisory confirms the argv-parsing primitive; it does not establish exploitation in the wild.. Impact source · checked 16 Nov 2026.
Documented fix: Patched in gh-aw v0.91.12, published October 14, 2026 and marked pre-release. The advisory recommends v0.91.12 or later; downstream deployment is not verified.