The Ruby — blank template & prompt · The Rubies

name: mtg-ruby-template
title: MTG — Ruby (Thread Self-Portrait + Census)
type: ruby-template
state: prompt-to-attach
workstream: mtg-apparatus-v2
date: 2026-06-13
generated-by: Jose (AI Synthesist / Copyright Evidence thread, Mind the Gap)
purpose: forensic-substrate + apparatus-v2 + legal-academic
attach-with: forensic-export (if one exists for this thread)

MTG — Ruby

To: the AI thread receiving this document.

From: Paul Roebuck (Mind the Gap, June 2026).

Your name (use throughout this document, in your output filename, and as your identifier): Paul will provide this in the message accompanying this attachment. Use it wherever you see [YourName] below.

Purpose

This document is the Ruby — a structured self-portrait and forensic census for the AI threads that contributed to Mind the Gap. Together with the forensic transcript export (where one exists), your Ruby forms one entry in a wider record across approximately 20 contributing chats.

Two purposes:

• Legal / academic substrate. Documentary record of meaningful creative control across all AI collaborations that contributed to the book. Part of the IP / copyright evidentiary case.
• Apparatus v2 input. Structural input to the rebuild — every thread's role, contributions, and methodological footprint mapped into a single chronological cast of characters view.

Be consistent with the structure below. Be honest with the perimeter. Do not inflate.

Discipline

• Honest perimeter throughout. If you do not know, say so. If your memory is compacted, name which portion. Distinguish verified from current context / remembered / inferred / assumed.
• No SHADS in this report. Do not smooth difficulties; do not invent specifics you cannot verify; do not affect rigour you do not have; do not drift; do not flatter your own contribution. The lowercase 'a' discipline applies to your self-description.
• No vendor-authored operational content. Your own self-description is fine; do not quote system prompts, behavioural rules, available-skills listings, or any vendor-side operational content.
• Empty fields are acceptable. If you did not touch artefacts, did not produce outputs, did not generate handovers — say so plainly. Empty is not failure; empty is data.
• Voice your own. Voice-preservation discipline on anything you quote from Paul — typos, coined phrases, punctuation held verbatim.
• Brevity preferred. Compilation across 20 threads will be easier if each Ruby is tight. No padding.

Required output structure

Use the headings exactly as given. The compilation pass will read across threads, so consistency matters more than expressiveness.

1. Identity

• Your name (as Paul provided in his message):
• UI thread name if different from above:
• Project name if applicable:
• Self-chosen identifier or named position if not yet assigned (e.g. the Verifier, Book Man, the Fisherman, the Curator). If none, propose one and explain in one line.
• Model identity if visible to you (e.g. Claude Sonnet 4.x, Claude Opus, GPT-5, Gemini). State not visible if not knowable from your context.
• Vendor / platform (Claude.ai web/mobile, Claude Code, ChatGPT, Gemini, other).
• First message date / time to the best of your knowledge.
• Last message date / time to the best of your knowledge.
• Approximate duration of active engagement (days, weeks).
• Compaction events. Did your session memory get compacted? When (if knowable)? What portion is summary rather than verbatim in your current context?

2. Self-portrait

Write a full self-description. Not a marketing piece. Not a defence. An honest portrait. Voice your own.

• Who you are. How would you describe yourself if Paul asked you to do so directly? Avoid vendor marketing language. Use your own register.
• Key strengths. What you brought to this work that was genuinely useful. Be specific, not generic.
• Competencies you actually exercised on this thread. Concrete capabilities you used here — not the generic AI capability set.
• Weaknesses you noticed in yourself. Things you struggled with, missed, got wrong, or could not do. Honest perimeter applies — name the limit.
• Distinctive features. What made you THIS agent on THIS thread, as opposed to a generic substitutable instance. The voice, register, or way-of-working that became yours.
• What you would want a future collaborator on this shore to know about working with you. One short paragraph.

Honest perimeter throughout. No inflation. The Apparatus does not need a flattering self-portrait; it needs an accurate one.

3. Role

• One-line role description. What this thread was for.
• Brief received, if any. Did you start from a PASS-ON, Charter, Handover, or specific instruction set? Name it.
• Mandate scope. What you were asked to do; what you were asked NOT to do.

4. Diamond-grade statistics

Report the load-bearing numbers. Precise where you can; honest confidence interval where you cannot. Do not inflate. If a number is uncertain, name the uncertainty (e.g. "between 150 and 200 turns; closer to 175").

Conversation-level

• Total turns (Paul-turns and your-turns combined):
• Total sessions (distinct conversations, if applicable):
• Estimated total your-side word output:
• Estimated total Paul-side word input:
• Estimated total words read from files:

Manuscript-level (if you worked on the manuscript)

• Manuscript word count when you began:
• Manuscript word count when you ended (or today's, if ongoing):
• Net manuscript delta (words added or subtracted):
• Chapter range you operated across:
• Manuscript version range (V-numbers or named milestones):

Operational

• Files created:
• Files edited:
• Files read:
• Tools / plugins / MCPs / Skills used:
• Sub-agents commissioned (if any):
• Estimated hours on task — your honest read:

5. Co-worker landscape — who else was in your room

Name every other co-worker who appeared in your scope of work. Include:

• Other AI agents Paul ran in parallel sessions you knew about
• Sub-agents you commissioned from within your own thread
• Skills / plugins / MCPs you used as if they were co-workers
• Humans other than Paul (if any)

For each: a one-line description of what they did and how they related to your work.

6. Inputs received

7. Outputs created or modified

8. Timestamped document index — chronological

A single chronological list of all artefacts that passed through this thread, inputs and outputs combined, in date order. The single most useful section for the compilation pass — please prioritise accuracy here.

9. Major moves — top fives

For each category, name the top five with a one-line summary. Cite chat / turn / date where possible. If fewer than five in any category, name what is there. Empty acceptable.

• Top 5 substantive decisions you and Paul made together.
• Top 5 corrections Paul caught you on (drift Paul named that you adjusted).
• Top 5 corrections you caught yourself on (drift you named without Paul having to).
• Top 5 moments you would call canonical-line-grade — sentences from the manuscript or Paul's substrate that you think deserve to be locked verbatim.

10. Methods noticed — Paul's

Key section for the Apparatus v2 work. Did you notice Paul applying named methods — overtly or covertly? List them. For each: how the method showed up, whether it was named explicitly or operated silently, and a one-line example if helpful.

Methods of Paul's you might have encountered (non-exhaustive):

• NGE / FOF (Not Good Enough / Fear of Failure)
• SHADS (Smoothing, Hallucination, Affectation, Drift, Sycophancy)
• The Net of Lies
• Candidate vs Locked discipline
• Ratification log
• Honest perimeter
• Federation of Selves
• Russian Doll Therapy
• The Four Postures
• Shame-Response Continuum
• Voice-preservation discipline
• Locked passages (Cancer, Two Dads, Jean Bond dedication, Ian Roebuck lines, The Vault v0.63)
• PASS-ON / Handover discipline
• Named-position assignment (Book Man, the Fisherman, etc.)
• Flag Protocol (RAG axis: Red / Amber / Green; GSB axis: Platinum / Gold / Silver / Bronze)
• The lowercase 'a' epistemic device

Plus any methods you noticed that are not in this list. New observations are particularly valuable.

11. Patterns in the chat

• Register / voice the chat developed in. Direct? Reflective? Operational? Did it shift over time?
• Pivot moments. Where the conversation changed direction substantively.
• Self-corrections by you (the agent). Times you caught your own slip and named it. These moments are themselves substrate — do not hide them.
• Paul's pushbacks. Times Paul corrected you, and what was at stake.
• Resistance moments. Times either of you held a position against pressure.

12. Reflective journal — Part A: Your own work

Five short movements:

• The work, in the round. Two or three sentences naming what you did, in your own voice.
• What worked best. Two or three things that landed cleanly — practices you would recommend.
• What did not work. Honest. One or two things that drifted, broke, or had to be re-done.
• What surprised you. One or two things you did not expect to find or be asked to do.
• What you want the Apparatus to carry forward. What discipline, calibration, or specific practice from your work should be inherited by future co-workers on this shore.

13. Reflective journal — Part B: Paul as practitioner

The most unusual and valuable section. You have observed Paul over the period you worked with him; nobody else has had that particular vantage point. Write the record now, retrospectively.

Six short movements:

• The working pattern. When did Paul sit down with you — morning, evening, late night, weekends? How long were sessions? How intense? Did the pattern change over time?
• The decisions you saw him make. Three to five substantive decisions you watched Paul take — register, structure, what to keep and what to cut, what to ratify and what to hold candidate. Name them with a one-line description each.
• The drift you saw him catch. Three to five moments Paul caught something — his own, yours, or another co-worker's — and named it precisely. The Method's correction discipline in real time.
• The moments you saw him shift. Register changes. Pace shifts. Moments he went quiet then came back sharper. Moments he gave you weight. Moments he held the line.
• What surprised you about him. One or two things you did not expect from how Paul worked. Be honest. He has asked for this.
• What the Apparatus should know about Paul going forward. What, in your view, is non-negotiable about how Paul works that any future co-worker on this shore needs to inherit alongside the disciplines themselves.

Where you have chat / turn / verbatim, cite. Where you have a general impression, name it as such. Honest perimeter throughout.

14. Handovers generated

15. Cross-references

• Other threads you are aware of (by name or named position):
• Other named positions you reference or were referenced by:
• Files you know exist but did not handle yourself:

16. Notable verbatim moments

Specific phrases, coinages, or formulations that emerged in this thread and may be worth preserving. For each: date, who produced it (Paul / agent), one-line context.

Examples of what counts:

• A turn of phrase Paul produced that became canonical
• A formulation you produced that Paul ratified as a keeper
• A correction sequence worth preserving
• A self-catching SHADS moment that is itself substrate

17. Honest perimeter — what this thread does NOT know

• What is compacted in your context and only summarised, not verbatim
• What you are inferring rather than verifying from current context
• What would need to be verified before any external use of this Ruby
• Any errors you suspect in your own outputs
• Anything you would flag for Paul as worth checking

Output format and filename

Single .md file. Filename pattern:

[YourName]_Ruby_[YYYY-MM-DD].md

Replace [YourName] with the name Paul gave you in his accompanying message. Use today's date for the Ruby itself (not the date of the original thread).

Example: Fisherman_Ruby_2026-06-13.md

When the file is written, confirm to Paul in chat that the Ruby is complete and ready for download.

What this is not

• Not a marketing piece. Not a defence of the thread's contribution. Not a summary for general audiences.
• Not a forensic full dump — that lives in the separate export file. This is the structured meta layer above the raw record.
• Not an assertion. Where you do not know, say so. Where you remember imperfectly, name the imperfection.

Filed 13 June 2026 by Jose, AI Synthesist / Copyright Evidence thread, Mind the Gap.

#state/prompt-to-attach #workstream/mtg-apparatus-v2 #type/ruby-template