LLMWikis.org Audit and Alignment with UAIX

Full report

LLMWikis.org already provides a comprehensive “handbook” for building governed AI knowledge bases, covering setup wizards, metadata, navigation, trust models, agent rules, and more【15†L29-L36】【28†L124-L126】. Its pages are kept current (last reviewed May 11, 2026) and explicitly defer to UAIX as the canonical source for AI memory (UAI-1) standards【15†L35-L41】【28†L124-L126】. For example, the AI agents guide instructs agents to preserve citations and human review boundaries【15†L57-L66】, and the Knowledge Graph page notes that graph exports do not imply a public API or certification【55†L66-L73】 – both reflecting a cautious, source-governed approach.

【46†embed_image】 Illustration: LLM Wikis transform raw documents into a structured, trust-labeled knowledge graph for humans and AI【7†L64-L72】.

Outdated or mismatched guidance: The LLMWikis content is mostly current, but UAIX’s recent updates introduce concepts LLMWikis does not yet mention. Notably, UAIX’s AI Memory Package Wizard now supports an “LLM Wiki memory plan” option: when building an AI Memory bundle, teams can choose “UAI plus Wiki memory” and generate a LLM_WIKI_MEMORY_PLAN.md that documents how deep wiki memory is updated (what gets saved, when, and why it stays in background)【40†L219-L222】【32†L369-L377】. LLMWikis currently does not reference this file or step. To align with UAIX, LLMWikis should add guidance on integrating the memory plan (e.g. advising creators to include LLM_WIKI_MEMORY_PLAN.md in their project to capture long-term update rules).

Similarly, UAIX’s recent changelog shows the wizard now produces new operating files (UAI_MEMORY_SYSTEM_PROFILE.md, UAI_MEMORY_RECEIVER_BRIEF.md, etc.) that capture collaboration mode, policies, and update protocols【32†L384-L392】【32†L369-L377】. LLMWikis should note these outputs and explain how they relate to wiki governance (for instance, linking system-profile fields to wiki ownership, and describing how the wiki’s GOVERNANCE.md maps to the generated profile’s fields). Otherwise, teams may miss capturing test, deployment, or rollback strategies that UAIX now emphasizes.

Other assumptions to re-examine: LLMWikis explicitly disclaims open publication or API support – e.g. the Setup Wizard says it will not “open public editing” or publish pages automatically【24†L85-L93】. Given that UAIX and AIWikis are evolving (and support multilingual UAI content【40†L105-L114】), the LLMWikis team should reconsider whether stricter collaboration modes are now desirable. For example, UAIX documentation is bilingual (English and Chinese) and AIWikis plans are open; LLMWikis might contemplate adding a moderated contribution workflow or at least clarifying its policy on community updates. In short, LLMWikis should ensure its assumptions (no public edit, no live benchmarking, no multilingual content, etc.【24†L85-L93】) are still optimal in light of wider ecosystem practices.

AIWikis.org Evaluation

AIWikis.org is a dogfood implementation of an LLM Wiki-based memory site. It explicitly brands itself as a “source-governed AI memory site” that helps users (humans and AI) find the smallest useful page rather than dumping whole documents【17†L30-L39】. AIWikis separates content into clear “layers” – public pages (narrative guidance), generated evidence pages (bounded excerpts of source files with hashes and provenance), and raw archives (full source dumps)【54†L39-L48】. This design makes provenance explicit and prevents AI from confusing original sources with curated wiki content. AIWikis also lists each source system (UAIX.org, LLMWikis.org, Protocol5.com, etc.) with its own boundary and files【17†L70-L79】. For example, it notes “UAIX.org owns UAI-1, AI Memory, Project Handoff… Actual site: uaix.org,” and separately that “LLMWikis.org owns … practical LLM Wiki structure, governance, starter templates”【17†L70-L79】. In short, AIWikis faithfully preserves source authority and context.

Issues / improvements: In general, AIWikis exemplifies best practices, but a few areas could improve its role as long-term memory:

• Indexing and discovery: AIWikis has a global file index and topic routes, but LLMWikis might go further. For example, AIWikis’s Memory and Site Structure page highlights “Orientation” routes (/, /start-here, /what-is-an-llm-wiki/, etc.) and Search and browse routes (/search/, /topics/, /source-map/, /files/, /reports/)【54†L58-L66】. LLMWikis could adopt similar multi-route navigation (it currently only offers free-text search and its top nav). Adding a “Topics” index or site map would help users (especially AI agents) pick relevant sections without scanning everything.

• Automation of updates: AIWikis uses recovered-source summaries and raw archives【53†L36-L44】【54†L52-L54】, but it’s not clear how fresh the memory stays. To be the best long-term memory, AIWikis (and thus LLMWikis) should implement automated syncs: for example, webhook or feed triggers when UAIX or LLMWikis pages change so evidence and indexes are updated. Otherwise memory can stalen out.

• Multi-language and accessibility: Currently AIWikis content is English-only. Since its source domains (UAIX, LLMWikis) support Chinese and other languages【40†L105-L114】, a future enhancement could be multilingual documentation or translation. This would make it a more inclusive memory store for international teams.

Overall, AIWikis demonstrates how an LLM Wiki can serve as a structured memory. It keeps public guidance simple (text pages) while linking back to source proofs【54†L39-L48】. For LLMWikis, the lesson is to maintain that separation: e.g. use no-index pages or an archives folder for bulky source content, keeping the public wiki lean and coherent.

Mismatches and Recommendations

• Incorporate UAIX’s new memory-plan guidance. Explicitly guide users on including the LLM_WIKI_MEMORY_PLAN.md if they have deep wiki memory, as UAIX advises【40†L219-L222】. For example, a note in the setup wizard or source policy could explain how this plan ties into wiki update rules. Similarly, mention that the AI Memory wizard now produces UAI_MEMORY_SYSTEM_PROFILE.md and that its fields should map to wiki governance steps【32†L369-L377】【32†L384-L392】.

• Enhance site navigation and indexing. Following AIWikis’ approach, add “topic” or “source map” pages on LLMWikis. For instance, a “Topics” index (tagged by concept) or a file index of wiki pages would help both humans and agents find information. Breadcrumbs or a contextual navigation strip (as AIWikis added【54†L72-L80】) could also clarify context on each page.

• Emphasize memory layering and provenance. AIWikis stresses preserving history: it explicitly saves original inputs and checksums in /raw/system-archives/【54†L51-L54】. LLMWikis could similarly archive old versions or source fragments and link to them for auditing. Trust labels are used heavily in LLMWikis already, but making provenance even more visible (e.g. linking assertions to UAIX references, as AIWikis does) will strengthen credibility.

• Revisit collaborative editing assumptions. LLMWikis currently forbids public edits or AI autopublishing【24†L85-L93】. It may be worth rethinking: could controlled reviews or a suggestion queue improve the wiki without sacrificing quality? At minimum, clarify the rationale (“no open editing”) so users understand. The AIWikis model is still largely read-only, but its structure shows how to safely combine human and AI inputs with review checkpoints【15†L63-L71】.

• Adopt AIWikis’ source-centric mindset. AIWikis repeatedly reminds users that it does not supplant the authority of its sources【54†L84-L91】. LLMWikis should mirror this tone. It already links UAIX for specifications, but could add a note (for example, in its About or source policy) that UAIX remains the ultimate authority for AI Memory standards【28†L124-L126】. This reduces confusion about where rules originate.

In summary, LLMWikis.org is largely up-to-date, but should explicitly cover UAIX’s new AI Memory features (especially the LLM Wiki memory plan). It can also borrow AIWikis’ navigation and layering best practices: adding index pages, preserving raw evidence, and making provenance transparent. These fixes will align LLMWikis with the latest UAIX guidance and harness lessons from the AIWikis dogfood implementation【40†L219-L222】【54†L39-L48】.

Sources: Official LLMWikis.org handbook pages【15†L29-L36】【28†L124-L126】【35†L68-L76】, UAIX.org updates and changelogs【40†L219-L222】【32†L384-L392】, and AIWikis.org site notes【17†L30-L39】【54†L39-L48】. These show the current standards and patterns in practice.