docs: prototype notice, disclaimers & ethics cross-links (#21)

Makes clear each repo is an **early-stage research prototype** (not a
product), adds the relevant warnings/disclaimers, explicitly welcomes
interest & collaboration, and links the ethics/UX context that pins down
the ethics side:

- [phronesis](https://github.com/hyperpolymath/phronesis) —
practical-wisdom language (normative substrate)
- [phronesiser](https://github.com/hyperpolymath/phronesiser) —
normative ethical constraints for AI agents
- [vexometer](https://github.com/hyperpolymath/vexometer) —
interaction-ethics / UX (Irritation Surface Analyser)
-
[palimpsest-license](https://github.com/hyperpolymath/palimpsest-license)
— ethical-use licence family

Adds `PROTOTYPE` notice + a README banner. Docs-only.

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>

Author: hyperpolymath

PROTOTYPE.md

@@ -0,0 +1,63 @@
+<!--
+SPDX-License-Identifier: CC-BY-SA-4.0
+SPDX-FileCopyrightText: 2025-2026 Jonathan D.A. Jewell <j.d.a.jewell@open.ac.uk>
+-->
+
+# ⚠️ Prototype status, disclaimers & how this fits the ethics work
+
+> **Aspasia is an early-stage research prototype, not a product.** It
+> explores one idea — an *independent neurosymbolic auditor* that
+> cross-checks statistical work (Octave + Prolog) — and shares it openly so
+> others can test it and push back.
+
+## What "prototype" means here
+
+- **Experimental.** Interfaces, the ABI/FFI surface, config formats and the
+  audit rules themselves are provisional and may change or be abandoned.
+- **Unproven.** The auditor has not itself been independently validated.
+  Coverage is partial; there are statistical situations it does not yet
+  understand.
+- **No warranty.** Provided "as is", with no warranty of any kind, to the
+  extent permitted by the licence. See `LICENSE` / `NOTICE`.
+
+## Auditing disclaimer (please read)
+
+Aspasia's whole job is to make you *more confident* in a result. That makes
+it dangerous to over-trust — a wrong audit is worse than no audit.
+
+- **The auditor is not an oracle.** A green verdict from Aspasia is *not*
+  proof that an analysis is correct, that the right test was chosen, or that
+  an interpretation is sound. A single unverified checker is itself a single
+  point of failure — the very problem it was built to address.
+- **Keep a human statistician in the loop.** Do not use Aspasia to sign off
+  consequential decisions (clinical, financial, legal, safety) unattended.
+- **Correcting the corrector cuts both ways.** Aspasia can be confidently
+  wrong for exactly the same reasons the LLM it audits can be. Treat its
+  output as a *second opinion to investigate*, not a ruling.
+
+## Interest and collaboration warmly welcome
+
+This is shared to *invite scrutiny*, not to advertise a finished thing.
+
+- Open an **issue** or **discussion** with counter-examples, cases it gets
+  wrong, or "an auditor should also check…".
+- Critique of the **ethics framing** (below) is especially valued.
+- See `CONTRIBUTING`, `CODE_OF_CONDUCT` and `EXHIBIT-A-ETHICAL-USE.txt`
+  before contributing.
+
+## Where the ethics thinking lives
+
+Auditing statistics is an *ethics* practice — it is about not misleading
+people with numbers, and about honest uncertainty. The estate develops that
+reasoning in dedicated sibling projects, and Aspasia defers to them:
+
+| Project | Role in the ethics picture |
+|---|---|
+| [**Phronesis**](https://github.com/hyperpolymath/phronesis) | A "practical wisdom" language — the estate's substrate for normative reasoning, distinct from a model's raw knowledge. |
+| [**Phronesiser**](https://github.com/hyperpolymath/phronesiser) | Adds **normative ethical constraints** to AI agents via Phronesis — the home for *what a responsible verdict owes the person receiving it*. |
+| [**Vexometer**](https://github.com/hyperpolymath/vexometer) | The **interaction-ethics / UX** side: an "Irritation Surface Analyser" for the human cost of tools. An auditor that nags or undermines confidence badly is itself a harm — this is how that cost gets named. |
+| [**Palimpsest License**](https://github.com/hyperpolymath/palimpsest-license) | The ethical-use licence family this repo ships under; see also this repo's own `EXHIBIT-A-ETHICAL-USE.txt` for the concrete ethical-use expectations. |
+
+For the ethics rationale specifically, start with **Phronesiser** (the
+*normative* side) and **Vexometer** (the *human-experience* side); this
+repo's `EXHIBIT-A-ETHICAL-USE.txt` grounds both in licence terms.

README.md

@@ -3,6 +3,18 @@ SPDX-License-Identifier: CC-BY-SA-4.0
 SPDX-FileCopyrightText: 2025-2026 Jonathan D.A. Jewell <j.d.a.jewell@open.ac.uk>
 -->
 
+> [!WARNING]
+> **Early-stage research prototype — not a product, and not an oracle.**
+> Interfaces and audit rules are experimental and may change or be abandoned.
+> A green verdict is a second opinion to investigate, **not** proof an
+> analysis is correct — keep a human in the loop. Interest, critique and
+> collaboration are warmly welcome.
+> See **[PROTOTYPE.md](PROTOTYPE.md)** for the full disclaimer and how this
+> connects to the ethics work —
+> [Phronesiser](https://github.com/hyperpolymath/phronesiser) (normative
+> constraints) and [Vexometer](https://github.com/hyperpolymath/vexometer)
+> (interaction ethics / UX).
+
 # The Name
 
 **Aspasia of Miletus** (~470-400 BCE) taught Socrates the art of