[Feature Request] Add a small "knowledge curation & memory failure modes" doc (docs only)

[onestardao]

Hi STORM team,

Thank you for releasing STORM. The idea of agents that curate long-term knowledge bases is very powerful and important for real-world use.

I have been working on 16-mode failure maps for RAG and memory-centric agents and recently contributed a robustness-related entry to Harvard MIMS Lab’s ToolUniverse. In systems like STORM, I see recurring classes of problems:

• knowledge entries are duplicated, inconsistent or partially conflicting
• the agent’s behavior diverges as the curated memory grows
• important context from earlier interactions is lost, while irrelevant details are preserved
• evaluation focuses on short-term answers instead of long-term knowledge quality

I would like to propose a small, docs-only feature.

Proposed feature

Add a short markdown page, for example:

storm_knowledge_and_memory_failure_modes.md

The page could:

1. List typical failure modes for knowledge curation and long-term memory:
   • duplication and inconsistency
   • drift as memory grows
   • forgetting important constraints.
2. For each, describe:
   • symptoms
   • likely causes (update policies, retrieval strategies, summarization)
   • basic diagnostics or sanity checks.
3. Provide a simple checklist for STORM users when they design new experiments or report issues.

Motivation

• STORM sits at the center of long-term memory and knowledge curation, where failures can be subtle.
• A small failure-mode guide would help practitioners reason about memory quality and debugging strategies.
• This is documentation only and should be straightforward to review and maintain.

If this seems in scope for the project, I would be glad to open a PR with a concise first draft.

Thank you for considering.

[onestardao]

Just a quick follow-up in case this issue got buried.

The idea was simply to add a short note about common knowledge retrieval failure modes (the 16-mode checklist) that people often run into when using retrieval-heavy benchmarks.

If it seems useful for STaRK users, I’d be happy to draft a small docs PR that fits the style of the repo.