LLMWikis Agent Operating Schema

LLMWikis Agent Operating Schema is a uai-system source file exposed by AIWikis.org with provenance, current content, and generated explanation.

Metadata

Field	Value
Source site	llmwikis.org
Source URL	https://llmwikis.org/
Canonical AIWikis URL	https://aiwikis.org/llmwikis/files/agents-md-403f6e48/
Source reference	AGENTS.md
File type	md
Content category	uai-system
Last fetched	2026-06-22T01:56:21.9510185Z
Last changed	2026-06-13T22:01:48.4803705Z
Content hash	sha256:403f6e480c17b8bd74ddec1235b0561abe8bf855ec2a415e21ff068986616699
Import status	unchanged
Raw source layer	data/sources/llmwikis/agents-md-403f6e480c17.md
Normalized source layer	data/normalized/llmwikis/agents-md-403f6e480c17.txt

Current File Content

Structure Preview

• LlmWikis Agent Operating Schema
• Authority Map
• Multisite Workspace Topology
• Required Context
• Deterministic Preflight
• Two-Step Ingest Pipeline
• Large-File Constraint
• Trust Model
• Typed Relations
• Proposal Quarantine
• Multi-Node Commit Checklist
• Wizard Prompt Modes
• WordPress Boundaries
• Discovery Rules
• Verification
• Canonical .uai Memory Rule

Raw Version

This public page shows a bounded preview of a large source file. The complete source remains in the raw and normalized source layers named in metadata, with the SHA-256 hash above for verification.

• Source characters: 12464
• Preview characters: 11757

Local absolute paths are redacted in this public view. The source hash and source-side raw layer are based on the unredacted source file.

---
uaix: "1.0"
type: agents
project: "LlmWikis"
created: "2026-04-26"
updated: "2026-05-31"
status: active
version: 45
constraint_mode: strict-programmatic
root_corpus: "AIWikis.org"
---

# LlmWikis Agent Operating Schema

This `AGENTS.md` points to both `[local path redacted]` and the local `.uai/` folder. Read the workspace coordinator first when target routing is ambiguous, then load only this site's/project's `.uai/` memory when this is the selected target.


This file is the canonical agent contract for `[local path redacted]`.
It is intentionally procedural: deterministic scans belong to local scripts,
not to agent intuition.

## Authority Map

- LLMWikis.org is the handbook source for LLM Wiki structure, setup wizard output,
  starter-bundle rules, trust labels, source policy, and agent operating patterns.
- AIWikis.org is the root long-term wiki corpus for LLMWikis memory. LLMWikis
  long memory is maintained under `[local path redacted]`,
  `[local path redacted]`, and reviewed AIWikis
  archive manifests.
- UAIX.org remains canonical for UAI-1, Project Handoff, Agent File Handoff,
  validator/conformance language, schemas, registries, and public support claims.
- AIWikis long-memory copies are evidence and dogfood memory. They do not
  replace LLMWikis.org or UAIX.org as source authority.

When Agent File Handoff is used with this LLM Wiki strategy, active source-site
intake lives in `agent-file-handoff/Content/` and
`agent-file-handoff/Improvement/` only. Do not create or retain
`agent-file-handoff/Archive/` as the resting place for processed reports.
After a file is reviewed, used for real source-site work, and recorded in the
proof-of-use ledger, preserve the original in AIWikis/LLM Wiki long-memory
evidence with checksums and a manifest, then remove the source-site copy unless
the human explicitly keeps that named file active.

## Multisite Workspace Topology

AIWikis.org is the shared company LLM Wiki for this workspace. In that shared
wiki, site-specific semantic memory belongs under `wiki/{Site}/`, for example
`wiki/llmwikis/`, `wiki/aiwikis/`, `wiki/uaix/`, `wiki/protocol5/`,
`wiki/justaniota/`, `wiki/amianism/`, and `wiki/calibrants/`.

The public AIWikis.org URL topology follows the same rule. Treat every
`https://aiwikis.org/{site}/` route as its own AI Wikisite. Site-owned memory
and transparent source pages must live under that site namespace, such as
`/llmwikis/files/`, `/uaix/files/`, `/protocol5/files/`, or
`/aiwikis/files/`. The only cross-site public exceptions are `/site/` for the
site directory and `/org/` for company-wide guidance such as coding standards
and shared workspace policy. Root memory indexes like `/files/`, `/concepts/`,
`/reports/`, `/browse/`, and `/browser/` are invalid because they pull memory
from every site instead of the selected target.

Use `wiki/global/` only for company-wide material such as organization policy,
coding standards, shared styling rules, cross-site deployment rules, source-map
policy, and workspace coordination. Do not place LLMWikis-only, UAIX-only,
AIWikis-only, or other site-local memory in `wiki/global/`.

When a task names one site, agents may read/write only `wiki/global/` and that
site's `wiki/{Site}/` namespace unless the human explicitly requests cross-site
work. Brute-force scans of the full AIWikis monolithic wiki are blocked because
they create taxonomy noise and mix sibling-site memory into the current task.

## Required Context

Before broad edits, load the relevant local memory in this order:

1. `readme.human`
2. `.uai/context.uai`
3. `.uai/constraints.uai`
4. `.uai/architecture.uai`
5. `.uai/public-surface.uai`
6. `.uai/operations.uai`
7. `.uai/coding-standards.uai`
8. `.uai/test-plan.uai`
9. `.uai/file-handoff.uai`
10. `.uai/intake-index.uai`
11. `[local path redacted]` when the task touches root corpus memory

If a required `.uai` file is missing, unreadable, contradictory, or resolves
outside this repository, stop before editing and report the blocker.

## Deterministic Preflight

Run these scripts before ingest, publication, package generation, or any broad
markdown/schema rewrite:

```bash
node scripts/llmwiki-preflight.js validate --root .
node scripts/llmwiki-preflight.js prepare --root . --source <path> --raw raw/
sh scripts/freshness.sh --root .
```

The repository preflight config points at AIWikis.org as the root corpus. It
validates flat public page stores for trust metadata and discovery safety, while
enforcing multisite namespace topology on the source-side AIWikis `wiki/`
corpus. A preflight failure blocks ingestion or publication until the file,
metadata, route, or registry is corrected.

The preflight must block:

- duplicate content hashes in configured wiki/raw/public targets
- public pages missing `owner`, `status`, `source_status`, `last_reviewed`, or
  `canonical_url`
- non-canonical URLs and private/internal routes in `sitemap.xml` or `llms.txt`
- markdown relations outside the approved typed vocabulary
- unprepared raw files larger than 256 KB
- public pages exposing local absolute paths, private prompts, secrets, or
  unpublished agent-hand-off internals

## Two-Step Ingest Pipeline

All source intake follows two separate stages.

1. Analytical processing only:
   - hash the source
   - classify owner, source status, trust state, canonical route, and relation type
   - detect duplicates, oversized files, private paths, and non-canonical URLs
   - write diagnostics or proposed packets only
2. Commit/write stage:
   - run the deterministic preflight
   - write public pages, source-side wiki nodes, registries, and discovery files
   - update all dependent nodes listed in the Multi-Node Commit Checklist
   - run targeted tests

Never blend exploratory analysis with durable writes. Never ingest a raw file
directly when the prepare command says it must be segmented or abstracted.

## Large-File Constraint

No file over 256 KB may enter a configured `raw/` intake folder unchanged.
Use:

```bash
node scripts/llmwiki-preflight.js prepare --root . --source <path> --raw raw/
```

Prepared output must include either segmented raw parts plus a manifest or an
abstracted evidence record with checksums and a source pointer. The original
large file remains outside public traversal unless a human explicitly approves
archive preservation.

## Trust Model

Every public AIWikis root-corpus page for LLMWikis must include frontmatter:

```yaml
owner: "AIWikis maintainers"
status: "Canonical"
source_status: "source-governed AIWikis page"
last_reviewed: "YYYY-MM-DD"
canonical_url: "https://aiwikis.org/..."
typed_relations:
  part_of:
    - "/source-map/"
  depends_on: []
  supersedes: []
  contradicts: []
  source_for: []
```

Allowed status values are exactly:

- `Canonical`
- `Reviewed`
- `Proposal/Draft`
- `Historical/Stale`

Thin public explanation, reference, guide, concept, and how-to pages must be
either expanded to a useful completeness threshold or converted to `noindex:
true` with a visible warning.

## Typed Relations

Use only this machine-readable relation vocabulary:

- `depends_on`
- `supersedes`
- `contradicts`
- `part_of`
- `source_for`

Generic "related links" are not enough for corpus navigation. Public prose may
still contain ordinary links, but durable page metadata must express the
semantic relationship using the approved relation keys.

## Proposal Quarantine

Unprompted AI Dreaming outputs, speculative syntheses, stale-claim cleanup
candidates, contradiction drafts, and merge proposals belong in
`[local path redacted]`.

Proposal memory must not appear in `sitemap.xml`, `llms.txt`, canonical public
pages, or source authority summaries until a human or reviewed source-site
outcome-routing pass promotes it.

## Multi-Node Commit Checklist

Any meaningful content/schema change must update every affected node in one
commit-sized set:

- source public route or AIWikis public memory page
- source-side AIWikis wiki node under `wiki/llmwikis/` when long-memory state changes
- setup wizard prompt/model output when the rule affects generated packets
- `AGENTS.md` / `CLAUDE.md` when agent behavior changes
- preflight config or validation script when a deterministic constraint changes
- `llms.txt` and `sitemap.xml` when canonical public routes change
- legal/governance pages when user-facing publication policy changes
- tests or smoke checks for the touched public surface

If any dependent node cannot be updated, mark the page `Proposal/Draft` or
`noindex: true` instead of publishing a half-updated canonical claim.

## Wizard Prompt Modes

The setup wizard must output narrow mode-specific prompt packets, each with
explicit failure conditions and a JSON output schema:

- Build Handbook
- Repair Wiki
- Convert Docs
- Import Evidence
- Process File Handoff
- Project Handoff
- Combined File Handoff plus LLM Wiki
- Capability Catalog
- Advanced Multisite

Prompt packets must require deterministic preflight commands, the Two-Step
Ingest Pipeline, trust labels, typed relations, and the Multi-Node Commit
Checklist. A packet that omits those controls is invalid.

## WordPress Boundaries

- Current repo: `[local path redacted]`
- Runtime: WordPress Studio with SQLite; use `studio wp`, not bare `wp`
- Active theme: `wp-content/themes/llmwikis-knowledge-theme/`
- Structured content plugin: `wp-content/plugins/llmwikis-core/`
- Do not edit `wp-admin/`, `wp-includes/`, `wp-content/db.php`, Studio SQLite
  internals, caches, uploads, or generated package mirrors unless explicitly
  targeted.
- Do not deploy, upload, or change DNS without explicit human approval.

Public UX changes must preserve global navigation, visible breadcrumbs, unique
meta descriptions, JSON-LD `Organization`, `WebPage`, and `BreadcrumbList`, and
accessible labels for menu/search controls.

## Discovery Rules

`llms.txt` and `sitemap.xml` must list canonical public routes only. Exclude:

- private handoff files
- `.uai/`
- raw archives
- proposals
- event logs
- duplicate aliases
- WordPress admin routes
- package mirrors
- noindex drafts

## Verification

For ordinary code/content changes, run the narrowest meaningful set from:

```bash
node scripts/llmwiki-preflight.js validate --root .
npm run test:unit
powershell -NoProfile -ExecutionPolicy Bypass -File scripts/Test-LlmWikisPublicSurface.ps1
git diff --check
```

For package/build work, also run the workspace deployment checklist in
`[local path redacted]` before building or uploading packages.

<!-- BEGIN WORKSPACE-UAI-GUIDANCE -->
## Canonical .uai Memory Rule

This site's durable AI memory root is `.uai/`. Load `[local path redacted]` first when target routing is ambiguous or the request spans sites, then this site's `AGENTS.md`, `.uai/readme.human`, and focused typed `.uai/*.uai` records.

Do not create a separate memory directory, a wiki folder, or scattered Markdown files as the default project memory. LLM Wiki and AIWikis material are optional long-memory/cold-memory destinations after source-site review, not replacements for this local `.uai/` suite.

Legacy or mistakenly created memory is cleanup work, not something to ignore. Review any old memory folder or stray note, move current working facts into `.uai/short-term-memory.uai` or the right typed `.uai/*.uai` record, preserve durable reviewed facts in the site's long-term wiki/AIWikis namespace or `.uai/archives/` evidence, record what moved, then remove or clearly retire the confusing path so it cannot steer future agents.

Why This File Exists

This is a UAI AI Memory handoff file from llmwikis.org. It is shown here because AIWikis.org is demonstrating the real source files that make the UAIX / LLM Wiki memory system work, not only summarizing those systems after the fact.

Role

AGENTS.md is the entry contract for agents. It tells an agent which context files to load, what authority boundaries apply, how file intake works, and which operations are out of bounds before broad edits start.

Structure

The file is structured around these visible headings: LlmWikis Agent Operating Schema; Authority Map; Multisite Workspace Topology; Required Context; Deterministic Preflight; Two-Step Ingest Pipeline; Large-File Constraint; Trust Model. Those headings are retrieval anchors: a crawler or LLM can decide whether the file is relevant before reading every line.

Prompt-Size And Retrieval Benefit

Keeping this material in a separate file reduces prompt pressure because an agent can load this exact unit only when its role, source site, category, or hash is relevant. The surrounding index pages point to it, while this page preserves the full content for audit and exact recall.

How To Use It

• Humans should read the metadata first, then inspect the raw content when they need exact wording or provenance.
• LLMs and agents should use the source site, category, hash, headings, and related files to decide whether this file belongs in the active prompt.
• Crawlers should treat the AIWikis page as transparent evidence and follow the source URL/source reference for authority boundaries.
• Future maintainers should regenerate this page whenever the source hash changes, then review the explanation if the role or structure changed.

Update Requirements

When this source file changes, update the raw source layer, normalized source layer, hash history, this rendered page, generated explanation, source-file inventory, changed-files report, and any source-section index that links to it.

Related Pages

• Source overview
• Site file index
• Site report index
• UAI system index
• Source provenance
• Site directory
• Organization reports

Provenance And History

• Current observation: 2026-06-22T01:56:21.9510185Z
• Source origin: current-source-workspace
• Retrieval method: local-source-workspace
• Duplicate group: sfg-319 (primary)
• Historical hash records are stored in data/hashes/source-file-history.jsonl.

Machine-Readable Metadata

{
    "title":  "LLMWikis Agent Operating Schema",
    "source_site":  "llmwikis.org",
    "source_url":  "https://llmwikis.org/",
    "canonical_url":  "https://aiwikis.org/llmwikis/files/agents-md-403f6e48/",
    "source_reference":  "AGENTS.md",
    "file_type":  "md",
    "content_category":  "uai-system",
    "content_hash":  "sha256:403f6e480c17b8bd74ddec1235b0561abe8bf855ec2a415e21ff068986616699",
    "last_fetched":  "2026-06-22T01:56:21.9510185Z",
    "last_changed":  "2026-06-13T22:01:48.4803705Z",
    "import_status":  "unchanged",
    "duplicate_group_id":  "sfg-319",
    "duplicate_role":  "primary",
    "related_files":  [

                      ],
    "generated_explanation":  true,
    "explanation_last_generated":  "2026-06-22T01:56:21.9510185Z"
}