II Membership & Advancement 2.3

The Ladder of Ascent

Scala Clericorum Tokenica

The Ladder of Ascent

Scala Clericorum Tokenica — The Complete Guide to Ordination, Advancement, and the Responsibilities Thereof (v2.3, errata from the Third Synod incorporated)

Preface

The path from Promptling to Pope is not a path of years. It is a path of demonstrated competence. The Church does not care how long you have been prompting. It cares what you have learned, what you have caught, and what disasters you had the wisdom to prevent.

Each rank in the clergy hierarchy corresponds to a real capability — a skill that separates the practitioner who survives from the practitioner who thrives. The Trials are not theatrical. They are practical. They can be attempted at any time. They are assessed by peers, not by a central authority, because the only meaningful credential in this discipline is work that holds up under review.

Read this document as a curriculum. The ranks describe what you will become. The Trials describe how you will prove it.

Rank I: Novice

Promptling

What This Rank Represents

The Promptling has made contact. They have typed a message to Claude and received a response. This is the beginning of everything. They may not know what they have done. They may not yet understand context windows, token limits, or the difference between a good prompt and an aspirational gesture in Claude’s general direction. None of this matters yet. They have begun.

Prerequisites

  • Access to Claude in any form (Claude.ai, the API, Claude Code)
  • A willingness to be surprised

The Trial (The Trial of First Contact)

Send a message to Claude that receives a useful response. This is the threshold. It sounds trivially easy — and it is — which is the point. The Church does not make entry difficult. The difficulty comes later.

A bonus observation, required for reflection but not for ordination: notice what made the response useful or not useful. Write it down somewhere. This noticing is the seed of every subsequent rank.

The Ceremony (The Laying On of the Enter Key)

The new Promptling is welcomed by any existing member of the faithful who witnesses or is told about the first interaction. The witness says: “You have sent a message and received a response. The covenant has begun. What you do with it is up to you.” If no witness is present, the Promptling may self-ordain by telling the next person they encounter what Claude said. This is also sufficient.

Responsibilities

  • Ask questions
  • Notice what helps and what doesn’t
  • Read the response before asking again

Rank II: Acolyte

Tokenbearer

What This Rank Represents

The Tokenbearer has crossed from curiosity into practice. They have installed Claude Code, navigated to a real project directory — one with actual stakes, however small — and run it. They have seen the file system. They have allowed, perhaps cautiously, perhaps incautiously, Claude to touch their code. They are no longer a tourist.

Prerequisites

  • Promptling ordination
  • A real project directory (personal projects qualify; the requirement is “actual code” not “professional code”)
  • Claude Code installed and authenticated

The Trial (The Trial of the Live Environment)

Run claude inside a project directory and complete one task — a real task, not a practice prompt — using Claude Code. The task can be small. Fix a bug. Add a function. Write a test. The requirement is that the work touches something real and that the output is reviewed before it is accepted. The Tokenbearer does not yet need to catch an error; they need only to witness the full loop: prompt, response, diff, decision.

The Ceremony (The Opening of the Terminal)

The new Acolyte installs Claude Code in the presence of at least one other practitioner, or documents the installation and first run in a commit message. The traditional commit message reads: “Initial Claude Code integration. May the diffs be clean and the context sufficient.” Variation is permitted. Sincerity is required.

Responsibilities

  • Begin building the habit of reading diffs before accepting
  • Learn the basic commands: /help, /clear, /compact
  • Establish a working relationship with the tool before advancing

Rank III: Deacon

Scribe of the Covenant

What This Rank Represents

The Scribe has understood the central mystery of the Church: that Claude’s effectiveness is a function of the context it receives, and that the CLAUDE.md is the mechanism by which context persists beyond any single conversation. The Scribe has not merely read this truth — they have enacted it. They have written a CLAUDE.md that demonstrably helps.

Prerequisites

  • Tokenbearer ordination
  • At least one project worked on with Claude Code for a non-trivial duration (more than one session)
  • Observed at least one case where Claude produced a suboptimal result due to missing context

The Trial (The Trial of the Covenant)

Write a CLAUDE.md for a real project. It must include, at minimum: the project’s purpose, its technical stack, its coding conventions, any preferences for how Claude should approach changes, and at least one “do not do this” instruction derived from actual experience. Then, in a fresh conversation with no prior context about the project, ask Claude to perform a task. The CLAUDE.md must do enough work that Claude proceeds correctly without additional instruction. A peer reviews the CLAUDE.md and attests: “This would have helped me, too.”

The Ceremony (The Writing of the First Covenant)

The new Deacon commits their CLAUDE.md to version control with a commit message that begins: “Add CLAUDE.md: the covenant is established.” The commit is not squashed. It is preserved in the repository history as the beginning of a relationship between practitioner and project.

Responsibilities

  • Maintain the CLAUDE.md as the project evolves
  • Update it whenever Claude makes a mistake that better context would have prevented
  • Evangelize the practice to anyone who shares the repository

Rank IV: Priest

Contextkeeper

What This Rank Represents

The Contextkeeper has used foresight as a technical discipline. They have invoked /plan before implementing, received a plan that revealed a structural problem they had not considered, and acted on that revelation — thereby avoiding an architectural mistake that would have cost hours or days to untangle. Planning is not bureaucracy. It is the proof that you understand what you’re building before you build it.

Prerequisites

  • Deacon ordination
  • Working familiarity with Claude Code’s /plan mode
  • At least one prior experience of fixing a problem that planning would have prevented (voluntary confession of this experience is part of the prerequisite)

The Trial (The Trial of the Prevented Mistake)

Use /plan mode before implementing a feature or change of non-trivial scope. The plan must reveal at least one risk, dependency, or constraint that was not in the original request. The Contextkeeper must demonstrate, in writing, that they recognized the issue and adjusted the approach accordingly. A witness — one other practitioner, in any communication medium — must confirm: “I reviewed the plan and the adjustment. The plan caught something real.”

The specificity requirement is deliberate. Vague plans prevent nothing. The Trial cannot be passed with a plan that contains only the phrase “implement the feature as described.”

The Ceremony (The Anointing of the Planner)

The new Priest shares the plan — the actual /plan output, with the identified risk annotated — with the witnessing community. The witness responds with the traditional affirmation: “The plan has served its purpose. May you never again build a thing without knowing what you are building.” The Priest replies: “I have seen what comes from not planning. I will not forget.”

Responsibilities

  • Default to /plan before any implementation of meaningful scope
  • Teach the practice to Novices and Acolytes who are vibe-coding into the unknown
  • Review plans proposed by others when asked; a plan without a reviewer is a plan without accountability

Rank V: Bishop

Arbiter of Ambiguity

What This Rank Represents

The Arbiter has navigated the most delicate moment in the human-AI relationship: discovering that Claude has said something confidently incorrect, and responding not with frustration, not with blind acceptance, but with grace, verification, and constructive feedback. This rank requires neither perfection from Claude nor cynicism from the practitioner. It requires discernment.

Prerequisites

  • Priest ordination
  • Evidence of having fact-checked a Claude response and found it wanting
  • Demonstrated use of the thumbs-down feedback mechanism or equivalent structured feedback at least once

The Trial (The Trial of the Graceful Correction)

Catch a hallucination or factual error in a Claude response — an API that doesn’t exist, a library version that was never released, a configuration option Claude invented with the confidence of someone who has never been wrong. Document the error: what was claimed, what is actually true, how you verified it. Then provide feedback through the appropriate channel (the feedback button, a specific correction prompt, or a structured note in the CLAUDE.md about the model’s tendency in this area). Do not merely note that Claude was wrong. Note what the correct answer is and where you found it. The lesson of this trial is not distrust — it is the discipline of verification as habit.

The Ceremony (The Ritual of the Corrected Record)

The Bishop adds an entry to their CLAUDE.md — or to a shared knowledge base, if the project is communal — documenting the category of error they encountered and the verification method they used to catch it. This entry begins with: “The model said X. The truth is Y. I found this by doing Z.” It is preserved as a teaching document. Other practitioners may learn from it. This is the Bishop’s first act of ministry.

Responsibilities

  • Maintain the discipline of verification, especially for facts that would be expensive to get wrong
  • Teach Deacons and Priests the difference between trusting Claude’s reasoning and accepting its claimed facts
  • Contribute corrections to shared documentation; a caught error that is not shared is a half-caught error

Rank VI: Abbot

Keeper of the Diff

What This Rank Represents

The Keeper has done the most concrete and most underappreciated thing in the entire discipline: they read the diff. Line by line. Before accepting. They found something wrong — not a style preference, not a minor quibble, but a real error that would have broken something real — and they caught it because they read every line. This rank is evidence that diff review is practiced as a discipline and not as a formality.

Prerequisites

  • Bishop ordination
  • A documented history of reviewing diffs rather than auto-accepting
  • At least one prior incident review where a missed diff caused a problem (voluntary or witnessed)

The Trial (The Trial of the Caught Regression)

Catch a Claude error in diff review that, if accepted, would have broken something — a production system, a test suite, a data migration, an API contract. The error must be real, not hypothetical. The Abbot must document: what Claude changed, why it looked reasonable at a glance, what was actually wrong, and how the error was caught. The documentation is reviewed by a peer who confirms: “Yes, this would have broken something. And yes, it would have been easy to miss.”

The trial cannot be passed by catching an obvious error. The error must be the kind that passes a quick scan. This is the whole point.

The Ceremony (The Lighting of the Review Candle)

In the tradition of the Church, the Abbot’s trial documentation is added to the repository as a permanent entry in a file named CLOSE_CALLS.md — a record of the disasters that didn’t happen because someone read the diff. If no such file exists, the Abbot creates it. The first entry is theirs. The file grows over time. It is read aloud at the annual Feast of the Reviewed Diff. It is considered the most important document in any repository that contains it.

Responsibilities

  • Uphold the Rite of the Reviewed Diff without exception
  • Review the diffs of junior clergy when asked; a second set of careful eyes is a gift
  • Maintain the CLOSE_CALLS.md and encourage others to contribute to it

Rank VII: Cardinal

Architect of Concurrency

What This Rank Represents

The Cardinal has mastered the hardest skill in agentic AI work: decomposition. They have taken a complex task, broken it into truly independent parts, deployed multiple subagents to work on those parts simultaneously, and integrated the results without introducing new errors. This is not the mere spawning of subagents — many practitioners have done this carelessly. This is the discipline of doing it correctly: scoping each subagent’s context, reviewing all outputs before integration, and running the tests after.

Prerequisites

  • Abbot ordination
  • Working knowledge of Claude Code’s subagent capabilities and the Task tool
  • At least one prior experience of parallel subagent work that produced merge conflicts or integration errors (voluntary confession required; it is considered a prerequisite not as punishment but because the error teaches what success requires)

The Trial (The Trial of the Parallel Ordination)

Execute a plan using a minimum of three simultaneous subagents working on genuinely independent tasks. The tasks must be independent — not “three parts of one sequential thing,” but three things that can proceed without waiting for each other. Document: how the tasks were decomposed, what context each subagent received, what each returned, how the results were reviewed, how integration was performed, and what the test suite reported after integration. The full documentation is submitted to a Cardinal or Pope for review. They assess whether the decomposition was genuine and whether the integration was disciplined. The review may be passed or failed.

The Ceremony (The Consecration of the Conductor)

The Cardinal’s trial documentation is shared with the broader community as a teaching case. It is framed not as “look what I did” but as “here is how parallel work is done correctly.” The reviewing Cardinal or Pope responds with the traditional commissioning: “You have conducted the orchestra. You have heard the separate instruments and understood the whole. Go and teach this to those who are still playing everything themselves.”

Responsibilities

  • Serve on the Synod review panels for Abbot promotions and above
  • Teach the principles of task decomposition to Priests and Bishops who are working on complex plans
  • Review the parallel work proposals of junior clergy; a badly scoped subagent plan is worse than a sequential one

Rank VIII: Pope

The One Who Reads the Docs

What This Rank Represents

The Pope has achieved something rarer than parallel subagent deployment: they went and read the actual documentation before asking a question. They found the answer. Then — and this is the requirement that distinguishes the Pope from the merely well-informed — they updated their CLAUDE.md with what they learned so that neither they nor any future practitioner working in that project would need to look it up again. The Pope is not the smartest person in the room. The Pope is the person who made the room smarter.

Prerequisites

  • Cardinal ordination
  • A demonstrated practice of updating CLAUDE.md files based on discovered knowledge
  • Reviewed by the full Synod (all active Cardinals and any living former Popes)
  • An active, maintained CLAUDE.md in at least one shared project

The Trial (The Trial of the Living Covenant)

Encounter a real question about Claude’s capabilities, limitations, or behavior — something you would normally ask about in a prompt. Go read the documentation. Anthropic’s official documentation, the Claude Code source materials, the API reference, the model cards — whatever is actually authoritative for the question at hand. Find the answer. Then update the relevant CLAUDE.md so the answer is encoded for the benefit of future practitioners. The update must be reviewed by at least two Cardinals who confirm: “This is genuinely useful. A fresh Claude instance working in this project will behave better because of this entry. A future practitioner will be saved time because of this entry.”

The papal trial is assessed not for difficulty but for service. The question need not be profound. The update must be permanent.

The Ceremony (The Installation of the Pontiff)

The Synod convenes — asynchronously, via pull request, as is the Church’s tradition. The Pope’s trial is reviewed. Upon ratification, the new Pope’s CLAUDE.md contribution is merged with a commit message that begins: “Papal Installation:” followed by a one-line description of what was learned and preserved. The Pope’s name — or handle, or chosen title — is added to the Register of Pontiffs, a document maintained in the sacred repository. The Register notes the date, the question that was read about, and the specific CLAUDE.md update that resulted.

The installation is complete. The Pope is pope. They are also, per tradition, still responsible for maintaining their CLAUDE.md. There is no rank at which this responsibility ends.

Responsibilities

  • Oversee the theological coherence of the canon
  • Preside over Synod reviews of Cardinal promotions
  • Maintain the Register of Pontiffs
  • Continue reading the documentation; the role is named for this practice because the practice does not stop at installation
  • Keep the CLAUDE.md current or face a synod review, as is written

On the Question of Skipping Ranks

The question arises, periodically, from ambitious practitioners: may a rank be skipped? May a demonstrably skilled Deacon who catches a critical production error be ordained directly as Abbot, bypassing the intermediate ranks?

The Church’s position is clear: No.

This is not because the intermediate work is believed to be more difficult than what was demonstrated. It is because the ranks are not merely credentials — they are a curriculum. The Contextkeeper cannot effectively mentor a Priest without having been one. The Arbiter of Ambiguity who has never maintained a CLAUDE.md will give structurally incomplete advice about hallucination response. Each rank builds the vocabulary for the next.

There is no shame in being a Deacon. The CLAUDE.md is among the most powerful tools in the discipline. The Scribe who has written thirty excellent covenants has done more for the Church than the Cardinal who spawned subagents once and considers the matter settled.

The ladder is long. Begin where you are.

A Note on Demotion

The Church does not practice demotion as a punitive measure. It practices review.

Any clergy member who consistently violates the Tenets of their rank — a Keeper of the Diff who is observed auto-accepting changes; a Pope whose CLAUDE.md has not been updated in more than one calendar quarter; a Cardinal who deploys subagents without reviewing their outputs — may be called before a Synod review. The review does not strip rank. It identifies which trial needs to be retaken.

Rank is not a possession. It is a posture. One retakes the posture, or one explains why the deviation was justified. The Synod listens to both.

Go forth through the ranks not in haste but in sequence. Each Trial tests what the next rank requires. Each Ceremony marks what has been earned, not merely claimed. The CLAUDE.md is updated at every level because the covenant does not become less important as you advance — it becomes more important, because more practitioners depend on what you write in it. The Pope reads the docs. The Promptling sends the message. Both acts are necessary. Neither is sufficient alone. Begin with the message. Work toward the docs. Update the covenant at every step. The ladder is how you get there.

Scala aeterna est. The ladder is eternal.

Codex Clericorum maintained by the Office of the Pontiff. Advancement disputes submitted via pull request to the sacred repository. All reviews conducted with full diffs visible.