Skip to content

Python: Proposal: official HOL Guard FunctionMiddleware example for protected tool calls #7833

Description

Proposal

Add an official Python example/docs integration showing HOL Guard as a concrete FunctionMiddleware enforcement layer for Agent Framework tool calls.

Agent Framework already has the right runtime boundary for this without a new generic API: FunctionMiddleware receives the FunctionInvocationContext before the wrapped function runs, and MiddlewareFailure provides an explicit fail-closed abort path.

The proposed official example would:

  • install and configure HOL Guard as the named security runtime;
  • evaluate the current function name and validated arguments in FunctionMiddleware.process(...) before call_next();
  • call call_next() only for an allow decision;
  • on deny, review-required, or Guard-unavailable/error paths, terminate before the wrapped tool executes, using Agent Framework's existing fail-closed middleware semantics;
  • include a focused test/example demonstrating allow => tool executes once and deny/failure => tool executes zero times.

This is intentionally a direct HOL Guard integration example, not a request for a new provider-neutral middleware abstraction or framework API. It would live in Agent Framework's official sample/docs surface so users can adopt the integration from the project itself.

If this placement sounds useful, I'm happy to send the smallest fork-first PR against main and keep it limited to the example/docs plus the minimal verification requested by maintainers.

Metadata

Metadata

Assignees

No one assigned

    Labels

    pythonUsage: [Issues, PRs], Target: PythontriageUsage: [Issues], Target: All issues that still need to be triaged

    Type

    No type

    Projects

    Status
    No status

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions