MCP Token Passthrough / Audience Confusion (Confused Deputy)
mcp-token-audience-confusion is a high severity check in the
ai category, written in python.
Its source is templates/ai/mcp/mcp-token-audience-confusion.py in
cert-x-gen-templates.
What it checks
Section titled “What it checks”ACTIVE check. Detects an MCP server that authenticates requests yet accepts a Bearer token whose audience (aud) is not scoped for it - the token passthrough / confused-deputy anti-pattern the June 2025 MCP spec explicitly forbids
Header
Section titled “Header”| Field | Value |
|---|---|
| Id | mcp-token-audience-confusion |
| Severity | high |
| Language | python |
| Category | ai |
| Author | Bugb Research |
| Template version | 1.0.0 |
| Confidence | 90 |
| CVSS | 8.1 |
| Weakness | CWE-287, CWE-441 |
| Tags | mcp, ai, agent, broken-auth, token-passthrough, audience-confusion, confused-deputy, oauth, intrusive, active, cwe-287, cwe-441 |
| Target kind | http |
| Oracle | property |
References
Section titled “References”Declared in the header. cxg parses @references and then discards it, and
nothing at scan time reads it, so this is the only place the links a
template cites are surfaced.
- https://modelcontextprotocol.io/specification/2025-06-18/basic/authorization
- https://modelcontextprotocol.io/specification/2025-06-18/basic/security_best_practices
- https://www.rfc-editor.org/rfc/rfc8707.html
- https://www.rfc-editor.org/rfc/rfc8693.html
- https://www.descope.com/blog/post/mcp-auth-spec
- https://www.practical-devsecops.com/glossary/token-passthrough-vulnerability/
- https://cwe.mitre.org/data/definitions/441.html
Run it
Section titled “Run it”To see what the copy on your machine says about itself, and to confirm it is installed at all:
cxg template info mcp-token-audience-confusionThe id it prints is the one to pass anywhere a template is selected. See
cxg template for the rest of
the subcommand, Scan a target for running
a scan, and
A match is not a finding for how
to read what comes back.
Related
Section titled “Related”- Template catalog is the whole corpus.
- ai is the rest of this category.
- Templates by weakness class groups checks by the defect they look for.
- Contribute a template is the route from a check you needed and could not find to a merged one.

