AAK-MCP-LANGGRAPH-CHECKPOINT-CVE-2026-71433-001: LangGraph Postgres and SQLite stores leak across tenant namespaces (postgres below 3.1.3, sqlite below 3.1.2)
- Severity
- MEDIUM
- Category
- Supply Chain (
supply-chain) - Scanner
agent_audit_kit/scanners/mcp_cve_pins_2026_07.py- Auto-fixable
- No.
- SARIF name
LangGraphCheckpointNamespaceLeak
What it detects
langgraph-checkpoint-postgres and langgraph-checkpoint-sqlite persist hierarchical namespaces as a dot-joined string, which twice let one tenant reach another's stored items. Before 3.1.1 a read was scoped by matching that string as a simple prefix pattern, so a read scoped to one namespace also matched a sibling whose flattened form shares the same leading characters, or a label holding unescaped pattern metacharacters, through an ordinary scoped search or list namespaces call (CVE-2026-71433, CVSS 5.3; the Postgres/SQLite sibling of the langgraph-checkpoint-mongodb checkpoint leak). Before postgres 3.1.3 and sqlite 3.1.2, PostgresStore, AsyncPostgresStore, SqliteStore and AsyncSqliteStore still joined labels with ".", while the base store validated labels only for put and aput, so a dotted label passed to get, search, delete, list_namespaces, their async forms or a direct batch resolved to another tenant's namespace and read, overwrote or deleted its items (CVE-2026-115883, CVSS 4.0 6.0, CWE-20/CWE-863, GHSA-6hmw-9432-67qq). InMemoryStore compares namespace tuples and is not affected. Treat a release below either floor, or an unpinned reference, as exposed.
How to fix
Upgrade langgraph-checkpoint-postgres to 3.1.3 or later and langgraph-checkpoint-sqlite to 3.1.2 or later, and pin them; both validate every namespace label before an operation reaches the store. Until then reject a namespace label holding a . before it reaches the store, for every operation, not only writes.
To confirm the finding is gone, report only this rule:
agent-audit-kit scan . --rules AAK-MCP-LANGGRAPH-CHECKPOINT-CVE-2026-71433-001
Limitations
Reads langgraph-checkpoint-postgres and langgraph-checkpoint-sqlite in requirements*.txt at any depth and, at the project root only, in pyproject.toml, Pipfile, package.json, .mcp.json, mcp.json, claude_desktop_config.json and the Pipfile.lock, poetry.lock, uv.lock, package-lock.json, pnpm-lock.yaml and yarn.lock lockfiles, taking the lowest version a range allows and reporting a reference that states no version. Each package is judged against its own floor, 3.1.3 for postgres and 3.1.2 for sqlite, so postgres 3.1.1 and 3.1.2 and sqlite 3.1.1, fixed for CVE-2026-71433, are reported for CVE-2026-115883. The base langgraph-checkpoint package, where the shared label check lives, is not read. The dependency is not read when it is declared in a sub-directory manifest (services/api/pyproject.toml) or in a lockfile format the reader does not parse (pdm.lock). A file over 2 MB, or a lockfile over 32 MiB, is not read. A declaration nested more than 64 levels deep in package.json, a TOML manifest or an MCP config (an MCP server more than 62) is still judged, and reported without a line; a lockfile entry keeps its line at any depth.
References
- CVE: CVE-2026-71433, CVE-2026-115883
- OWASP Agentic Top 10: ASI03 (Identity and Privilege Abuse), ASI04 (Agentic Supply Chain Vulnerabilities)
- Adversa AI MCP Top 25 (AAK-defined ids): ADV-SUPPLY-01 (Runtime Package Fetch)
- CSA AI Controls Matrix (AICM): STA-08, AIS-07, DSP-04
All rules. This page is generated from the rule registry when the documentation is built; to change it, change the rule in agent_audit_kit/rules/builtin.py. To give the rule a hand-written write-up, add docs/rules/AAK-MCP-LANGGRAPH-CHECKPOINT-CVE-2026-71433-001.md: the page then shows that text, with the registry reference appended under it.