Skip to content

docs: add guarded agent memory cookbook - #313

Open
vgudur-dev wants to merge 1 commit into
deepset-ai:mainfrom
vgudur-dev:contrib/amg-memory-poisoning-notebook
Open

docs: add guarded agent memory cookbook#313
vgudur-dev wants to merge 1 commit into
deepset-ai:mainfrom
vgudur-dev:contrib/amg-memory-poisoning-notebook

Conversation

@vgudur-dev

Copy link
Copy Markdown

Summary

This adds a focused, no-API-key notebook showing a guarded durable-memory boundary for Haystack agents. A benign external observation reaches Haystack's InMemoryDocumentStore only after inspection by OWASP Agent Memory Guard (AMG). A representative indirect prompt-injection payload is blocked before write_documents() is called, and the resulting security event remains available for logging or investigation.

The accompanying index.toml entry categorizes the notebook under Agents, Memory, and Guardrails.

Validation

The notebook scenario was exercised against public haystack-ai==3.0.0 and agent-memory-guard==0.3.0. It confirms that a benign observation persists, while the poisoned tool-output fixture raises AMG's strict-policy violation, emits a prompt_injection event with EXTERNAL_TOOL provenance, and never reaches the document store.

Scope

This is a small, deterministic example notebook. It uses Haystack's in-memory document store, requires no model provider or API key, and does not claim complete agent-security coverage.

@vgudur-dev
vgudur-dev requested a review from a team as a code owner August 13, 2026 05:42
@review-notebook-app

Copy link
Copy Markdown

Check out this pull request on  ReviewNB

See visual diffs & provide feedback on Jupyter Notebooks.


Powered by ReviewNB

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant