> ## Documentation Index
> Fetch the complete documentation index at: https://docs.wednesdayai.dev/llms.txt
> Use this file to discover all available pages before exploring further.

# Product naming

> The WednesdayAI/openclaw naming contract for plugin and core contributors: exact case map, KEEP vs REPLACE classes, prompt markers, and guard tests.

# Product naming

Model-visible agent context uses the **WednesdayAI** brand. Operator identifiers, disk paths, env vars, and SDK package names stay **openclaw**. This page is the contract for code and tests that touch either side.

## Case map (exact)

| Form                                    | Replace with  |
| --------------------------------------- | ------------- |
| `openclaw` (lowercase, prose/teaching)  | `wednesdayai` |
| `OPENCLAW` (uppercase, prose/teaching)  | `WEDNESDAYAI` |
| `OpenClaw` (mixed case, prose/teaching) | `WednesdayAI` |

Do not invent other spellings (`Wednesday AI`, `wednesday-ai`, `Openclaw`, …).

## KEEP classes (never blind-rename)

Even when a string is model-visible or adjacent to REPLACE text, keep these byte-identical:

* `OPENCLAW_*` env vars
* `~/.openclaw/`, `openclaw.json`, service/unit names, sockets
* TypeScript identifiers (`OpenClawConfig`, `OpenClawPluginApi`, …)
* Import specifier `openclaw/plugin-sdk` (the `wednesdayai/plugin-sdk` dual alias is also valid where published)
* Manifest filename `openclaw.plugin.json` and the legacy `openclaw` manifest key
* `__OPENCLAW_REDACTED__`, `openclaw.inbound_meta.v1`, `provider: "openclaw"`
* npm package name `wednesdayai` with dual `bin`: `wednesdayai` **and** `openclaw` (both point at `openclaw.mjs` — the alias is permanent)

When a line mixes KEEP and REPLACE (example: `profile="openclaw"` next to "openclaw-managed browser" prose), change **only** the prose token.

## Surfaces that must say WednesdayAI

* System prompt identity and section headings, docs block, tool blurbs
* Subagent / announce / group-intro strings the model reads
* Tool error and status text reused into model context (session status version line included)
* Bundled skill descriptions/bodies and workspace templates
* Model-taught docs base URL `https://docs.wednesdayai.dev`
* Model-taught CLI spelling: primary binary `wednesdayai` (`openclaw` remains a permanent alias at the process boundary)

## Prompt markers (lockstep)

```typescript theme={"dark"}
// src/agents/system-prompt.ts
export const SYSTEM_PROMPT_CACHE_BOUNDARY = "<!-- WEDNESDAYAI_CACHE_BOUNDARY -->";
export const SYSTEM_PROMPT_PROJECT_CONTEXT_END = "<!-- WEDNESDAYAI_PROJECT_CONTEXT_END -->";
```

The system-prompt report parser **must** import these constants — never re-hardcode marker literals. Assembler and parser are same-binary pairs (ADR 0051).

## Plugin authoring rules

```typescript theme={"dark"}
// ✅ Public import surface — stable path, WednesdayAI-agnostic
import type { OpenClawPluginApi } from "openclaw/plugin-sdk";
```

* Import from `openclaw/plugin-sdk` (or the published `wednesdayai/plugin-sdk` dual alias), never from relative paths into `src/`.
* Your manifest filename stays `openclaw.plugin.json`; either `openclaw` or `wednesdayai` works as the `package.json` manifest key.
* User-facing prose in your plugin (descriptions, reply text the model echoes) should say WednesdayAI; config keys and file paths you invent should follow the operator convention.

## Guard tests

| Test                                           | Role                                                                               |
| ---------------------------------------------- | ---------------------------------------------------------------------------------- |
| `src/agents/system-prompt.branding.test.ts`    | Assembled prompt contains WednesdayAI identity, no stale `running inside OpenClaw` |
| `src/agents/context-branding-grep.test.ts`     | Repo sweep gate for high-risk model-context paths                                  |
| `src/auto-reply/reply/groups.branding.test.ts` | Group-intro branding assembly                                                      |
| `scripts/docs-surface-product-naming.test.ts`  | docs-surface.yaml product-naming mapping                                           |

Update colocated string assertions in the **same change** as a source rename. Prefer at least one test that builds the real prompt (or real tool result) rather than only checking a helper's return value.

## What not to do

```typescript theme={"dark"}
// ❌ Re-hardcode markers in the report parser
const end = "<!-- WEDNESDAYAI_CACHE_BOUNDARY -->\n";

// ❌ Rename env / state dir "for consistency"
process.env.WEDNESDAYAI_STATE_DIR;

// ❌ Teach the model docs.openclaw.ai or "openclaw gateway" in new prompt copy
```

```typescript theme={"dark"}
// ✅ Import shared marker constants
import { SYSTEM_PROMPT_CACHE_BOUNDARY } from "./system-prompt.js";

// ✅ KEEP operator identity
process.env.OPENCLAW_STATE_DIR;
// ~/.openclaw/openclaw.json
```

## Related

* [Product naming (admin)](/admin/product-naming) — the operator guide
* [What is WednesdayAI?](/users/what-is-wednesdayai) — user summary
* [Plugin manifest](/developers/plugins/manifest) — manifest key rules