Proposal: add optional Omega Walls guard around Haystack agent tools

[anviren]

Hi Haystack team,

This repo looks like a good practical example of a Haystack RAG agent with customizable tools.

One thing that could be useful as an optional example:

how to add a runtime trust boundary when the agent reads external content and can call tools.

The risk is not only “security” in the narrow sense. It is also reliability:

retrieved content, web pages, PDFs, tickets, or tool outputs can enter the same context path as trusted instructions. Once tools are available, that can make the agent workflow harder to reason about.

I’m working on Omega Walls, an open-source trust-boundary layer for RAG and agent pipelines.

It already has a Haystack adapter:

from omega.integrations import OmegaHaystackGuard

guard = OmegaHaystackGuard(profile="quickstart")

pipeline = guard.wrap_pipeline(
    pipeline,
    component_name="omega_guard_component",
)

safe_tool = guard.wrap_tool("network_post", network_post_fn)

Install:

pip install "omega-walls[integrations]"

Possible contribution:

• add an optional omega_walls_guard example
• wrap the Haystack pipeline
• wrap one example tool from tools.py
• demonstrate blocked tool execution on unsafe/malformed input
• keep the default demo unchanged

Links:
https://github.com/synqratech/omega-walls

https://pypi.org/project/omega-walls/

If this fits the scope of the repo, I can open a small PR.