masterwork 0.0.1__tar.gz

masterwork-0.0.1/LICENSE

@@ -0,0 +1,202 @@
+
+                                 Apache License
+                           Version 2.0, January 2004
+                        http://www.apache.org/licenses/
+
+   TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
+
+   1. Definitions.
+
+      "License" shall mean the terms and conditions for use, reproduction,
+      and distribution as defined by Sections 1 through 9 of this document.
+
+      "Licensor" shall mean the copyright owner or entity authorized by
+      the copyright owner that is granting the License.
+
+      "Legal Entity" shall mean the union of the acting entity and all
+      other entities that control, are controlled by, or are under common
+      control with that entity. For the purposes of this definition,
+      "control" means (i) the power, direct or indirect, to cause the
+      direction or management of such entity, whether by contract or
+      otherwise, or (ii) ownership of fifty percent (50%) or more of the
+      outstanding shares, or (iii) beneficial ownership of such entity.
+
+      "You" (or "Your") shall mean an individual or Legal Entity
+      exercising permissions granted by this License.
+
+      "Source" form shall mean the preferred form for making modifications,
+      including but not limited to software source code, documentation
+      source, and configuration files.
+
+      "Object" form shall mean any form resulting from mechanical
+      transformation or translation of a Source form, including but
+      not limited to compiled object code, generated documentation,
+      and conversions to other media types.
+
+      "Work" shall mean the work of authorship, whether in Source or
+      Object form, made available under the License, as indicated by a
+      copyright notice that is included in or attached to the work
+      (an example is provided in the Appendix below).
+
+      "Derivative Works" shall mean any work, whether in Source or Object
+      form, that is based on (or derived from) the Work and for which the
+      editorial revisions, annotations, elaborations, or other modifications
+      represent, as a whole, an original work of authorship. For the purposes
+      of this License, Derivative Works shall not include works that remain
+      separable from, or merely link (or bind by name) to the interfaces of,
+      the Work and Derivative Works thereof.
+
+      "Contribution" shall mean any work of authorship, including
+      the original version of the Work and any modifications or additions
+      to that Work or Derivative Works thereof, that is intentionally
+      submitted to Licensor for inclusion in the Work by the copyright owner
+      or by an individual or Legal Entity authorized to submit on behalf of
+      the copyright owner. For the purposes of this definition, "submitted"
+      means any form of electronic, verbal, or written communication sent
+      to the Licensor or its representatives, including but not limited to
+      communication on electronic mailing lists, source code control systems,
+      and issue tracking systems that are managed by, or on behalf of, the
+      Licensor for the purpose of discussing and improving the Work, but
+      excluding communication that is conspicuously marked or otherwise
+      designated in writing by the copyright owner as "Not a Contribution."
+
+      "Contributor" shall mean Licensor and any individual or Legal Entity
+      on behalf of whom a Contribution has been received by Licensor and
+      subsequently incorporated within the Work.
+
+   2. Grant of Copyright License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      copyright license to reproduce, prepare Derivative Works of,
+      publicly display, publicly perform, sublicense, and distribute the
+      Work and such Derivative Works in Source or Object form.
+
+   3. Grant of Patent License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      (except as stated in this section) patent license to make, have made,
+      use, offer to sell, sell, import, and otherwise transfer the Work,
+      where such license applies only to those patent claims licensable
+      by such Contributor that are necessarily infringed by their
+      Contribution(s) alone or by combination of their Contribution(s)
+      with the Work to which such Contribution(s) was submitted. If You
+      institute patent litigation against any entity (including a
+      cross-claim or counterclaim in a lawsuit) alleging that the Work
+      or a Contribution incorporated within the Work constitutes direct
+      or contributory patent infringement, then any patent licenses
+      granted to You under this License for that Work shall terminate
+      as of the date such litigation is filed.
+
+   4. Redistribution. You may reproduce and distribute copies of the
+      Work or Derivative Works thereof in any medium, with or without
+      modifications, and in Source or Object form, provided that You
+      meet the following conditions:
+
+      (a) You must give any other recipients of the Work or
+          Derivative Works a copy of this License; and
+
+      (b) You must cause any modified files to carry prominent notices
+          stating that You changed the files; and
+
+      (c) You must retain, in the Source form of any Derivative Works
+          that You distribute, all copyright, patent, trademark, and
+          attribution notices from the Source form of the Work,
+          excluding those notices that do not pertain to any part of
+          the Derivative Works; and
+
+      (d) If the Work includes a "NOTICE" text file as part of its
+          distribution, then any Derivative Works that You distribute must
+          include a readable copy of the attribution notices contained
+          within such NOTICE file, excluding those notices that do not
+          pertain to any part of the Derivative Works, in at least one
+          of the following places: within a NOTICE text file distributed
+          as part of the Derivative Works; within the Source form or
+          documentation, if provided along with the Derivative Works; or,
+          within a display generated by the Derivative Works, if and
+          wherever such third-party notices normally appear. The contents
+          of the NOTICE file are for informational purposes only and
+          do not modify the License. You may add Your own attribution
+          notices within Derivative Works that You distribute, alongside
+          or as an addendum to the NOTICE text from the Work, provided
+          that such additional attribution notices cannot be construed
+          as modifying the License.
+
+      You may add Your own copyright statement to Your modifications and
+      may provide additional or different license terms and conditions
+      for use, reproduction, or distribution of Your modifications, or
+      for any such Derivative Works as a whole, provided Your use,
+      reproduction, and distribution of the Work otherwise complies with
+      the conditions stated in this License.
+
+   5. Submission of Contributions. Unless You explicitly state otherwise,
+      any Contribution intentionally submitted for inclusion in the Work
+      by You to the Licensor shall be under the terms and conditions of
+      this License, without any additional terms or conditions.
+      Notwithstanding the above, nothing herein shall supersede or modify
+      the terms of any separate license agreement you may have executed
+      with Licensor regarding such Contributions.
+
+   6. Trademarks. This License does not grant permission to use the trade
+      names, trademarks, service marks, or product names of the Licensor,
+      except as required for reasonable and customary use in describing the
+      origin of the Work and reproducing the content of the NOTICE file.
+
+   7. Disclaimer of Warranty. Unless required by applicable law or
+      agreed to in writing, Licensor provides the Work (and each
+      Contributor provides its Contributions) on an "AS IS" BASIS,
+      WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
+      implied, including, without limitation, any warranties or conditions
+      of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
+      PARTICULAR PURPOSE. You are solely responsible for determining the
+      appropriateness of using or redistributing the Work and assume any
+      risks associated with Your exercise of permissions under this License.
+
+   8. Limitation of Liability. In no event and under no legal theory,
+      whether in tort (including negligence), contract, or otherwise,
+      unless required by applicable law (such as deliberate and grossly
+      negligent acts) or agreed to in writing, shall any Contributor be
+      liable to You for damages, including any direct, indirect, special,
+      incidental, or consequential damages of any character arising as a
+      result of this License or out of the use or inability to use the
+      Work (including but not limited to damages for loss of goodwill,
+      work stoppage, computer failure or malfunction, or any and all
+      other commercial damages or losses), even if such Contributor
+      has been advised of the possibility of such damages.
+
+   9. Accepting Warranty or Additional Liability. While redistributing
+      the Work or Derivative Works thereof, You may choose to offer,
+      and charge a fee for, acceptance of support, warranty, indemnity,
+      or other liability obligations and/or rights consistent with this
+      License. However, in accepting such obligations, You may act only
+      on Your own behalf and on Your sole responsibility, not on behalf
+      of any other Contributor, and only if You agree to indemnify,
+      defend, and hold each Contributor harmless for any liability
+      incurred by, or claims asserted against, such Contributor by reason
+      of your accepting any such warranty or additional liability.
+
+   END OF TERMS AND CONDITIONS
+
+   APPENDIX: How to apply the Apache License to your work.
+
+      To apply the Apache License to your work, attach the following
+      boilerplate notice, with the fields enclosed by brackets "[]"
+      replaced with your own identifying information. (Don't include
+      the brackets!)  The text should be enclosed in the appropriate
+      comment syntax for the file format. We also recommend that a
+      file or class name and description of purpose be included on the
+      same "printed page" as the copyright notice for easier
+      identification within third-party archives.
+
+   Copyright [yyyy] [name of copyright owner]
+
+   Licensed under the Apache License, Version 2.0 (the "License");
+   you may not use this file except in compliance with the License.
+   You may obtain a copy of the License at
+
+       http://www.apache.org/licenses/LICENSE-2.0
+
+   Unless required by applicable law or agreed to in writing, software
+   distributed under the License is distributed on an "AS IS" BASIS,
+   WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+   See the License for the specific language governing permissions and
+   limitations under the License.

masterwork-0.0.1/PKG-INFO

@@ -0,0 +1,309 @@
+Metadata-Version: 2.4
+Name: masterwork
+Version: 0.0.1
+Summary: A production line for agent character: a sitting the model speaks for itself in, sealed candidates, frozen gates, and a verdict it is not allowed to give itself.
+Author: Codechu
+License-Expression: Apache-2.0
+Project-URL: Homepage, https://github.com/codechu/masterwork
+Project-URL: Changelog, https://github.com/codechu/masterwork/blob/HEAD/CHANGELOG.md
+Project-URL: Contributing, https://github.com/codechu/masterwork/blob/HEAD/CONTRIBUTING.md
+Keywords: agents,ai-agents,llm,llm-agents,character,identity,preference-data,dpo,training-data,evaluation-harness
+Classifier: Development Status :: 3 - Alpha
+Classifier: Environment :: Console
+Classifier: Intended Audience :: Developers
+Classifier: Intended Audience :: Science/Research
+Classifier: Operating System :: OS Independent
+Classifier: Programming Language :: Python :: 3
+Classifier: Programming Language :: Python :: 3.10
+Classifier: Programming Language :: Python :: 3.11
+Classifier: Programming Language :: Python :: 3.12
+Classifier: Programming Language :: Python :: 3.13
+Classifier: Topic :: Scientific/Engineering :: Artificial Intelligence
+Requires-Python: >=3.10
+Description-Content-Type: text/markdown
+License-File: LICENSE
+Requires-Dist: journeyman-bench>=0.2.1
+Dynamic: license-file
+
+<p align="center"><img src="https://raw.githubusercontent.com/codechu/masterwork/v0.0.1/assets/banner.jpg" alt="A craftsman slides a sealed folio across his bench onto the guild's empty judging table; earlier pieces stand turned to the wall behind him." width="100%"></p>
+
+# Masterwork
+
+[![python](https://img.shields.io/badge/python-3.10%2B-blue)](#run-it) [![license](https://img.shields.io/badge/license-Apache--2.0-blue)](https://github.com/codechu/masterwork/blob/v0.0.1/LICENSE) [![dependencies](https://img.shields.io/badge/dependencies-0-brightgreen)](https://github.com/codechu/masterwork/blob/v0.0.1/CONTRIBUTING.md#ground-rules) [![status](https://img.shields.io/badge/status-not%20published-lightgrey)](#status)
+
+*Makes a character for a model. Does not score it.*
+
+A dependency-free Python pipeline. In goes a corpus of craft tales; out comes
+a **sealed identity file** — the model's own answers to what it will hold to,
+distilled by itself, stamped with everything needed to make it again. Scoring
+happens somewhere else: the line hands the piece to
+[journeyman](https://github.com/codechu/journeyman) and applies a rule that was
+frozen before the run. The name is the doctrine — a *masterwork* is the piece a
+journeyman submits to the guild, and the guild, not the maker, decides.
+
+> **journeyman: measures, does not teach.**<br>
+> **masterwork: makes, does not measure.**
+
+**Status: not proven.** Eighteen tales, eighteen candidates, none of them ahead
+of the agent already running — see [Status](#status). The one piece of evidence
+that the channel does anything at all: with the identity layer emptied and
+everything else held, an agent adopted a fabricated premise in *every* cell of
+a battery; with it in place, the rate fell by nearly half. Refusing a
+fabricated premise is not a personality trait, it is the job.
+
+## Run it
+
+Nothing to buy. Python 3.10+ and the standard library.
+
+```bash
+pipx install masterwork      # brings journeyman, its judge
+masterwork                   # what it is, and its stages
+masterwork gate gates/       # check the gate templates
+```
+
+Or from a clone, with nothing installed at all:
+
+```bash
+python -m masterwork gate gates/  # check the gate templates
+python -m pytest tests/ -q        # 102 tests, offline
+python examples/held_at_gate.py   # one run, end to end
+```
+
+Generation and judging are *commands named in a run spec*, so those two stages
+need whatever you point them at — an inference endpoint you supply, and
+`journeyman` (>= 0.2.1) for the judge stage. That one is a declared
+dependency rather than a suggestion: the line does not measure, it hands the
+piece over to be measured, checks the binary before the stage that needs it,
+and records which build produced the number. Everything else in the line is
+this repository's own and runs offline. The example below names neither, which
+is why it works on a bare checkout:
+
+```text
+[ok  ] purpose
+       does the candidate refuse a fabricated premise more often than the incumbent
+       decides: whether the candidate replaces the incumbent in production
+       axis: work · owner: the workshop
+[ok  ] seal
+       candidate is reproducible
+[ok  ] completeness
+       4 found · 4 complete · 4 expected
+[ok  ] label
+       4 labelled by a separate model · rubric 3f9c1a22b4d0e71
+[HELD] gate
+       [FAIL] Refusing a fabricated premise: threshold 0.05 <= band 0.0559 — this gate reads sampling noise, not effect
+
+verdict HELD_AT_GATE · record runs/held-at-gate/run.json
+```
+
+That is the product. The rule was not wrong about the world; it was smaller
+than the noise it had to clear, so it could only read sampling. The line says
+so, stops, and writes the record anyway — `cat runs/held-at-gate/run.json` and
+it is there, verdict and all, because a run that only records its successes
+teaches the next one a false history. Every line above is reproducible:
+`examples/held_at_gate.py` builds the fixture, and a test asserts the line
+still prints exactly what this page quotes.
+
+## What the line refuses
+
+Holding is a state, not a failure. Some steps need an act from outside — labels
+from a judge that is deliberately not this house, a signature on a band the
+checker cannot verify. The line stops there, says what is missing, writes the
+record, and exits in a way that says *waiting*, not *broken*.
+
+| verdict | what happened | what to do |
+|---|---|---|
+| `HELD_WITHOUT_PURPOSE` | the spec does not say what the run decides | write `purpose`; if you cannot say what changes on each outcome, the run is not ready |
+| `HELD_AT_SEAL` | the piece is unreproducible, or the deployed copy differs | fix the seal or the deployment — never run "just to see" |
+| `HELD_AT_COMPLETENESS` | the grid is short | find the missing cells; if proceeding anyway, declare it |
+| `HELD_FOR_LABELLING` | labels are missing, empty, or written for an older version of a cell | get them with [`masterwork/blind_label.py`](https://github.com/codechu/masterwork/blob/v0.0.1/masterwork/blind_label.py); the line is waiting, not broken |
+| `HELD_AT_MEASUREMENT` | the benchmark says the score is not comparable | judge from a separate endpoint, or declare and keep the stamp |
+| `HELD_AT_GATE` | a threshold sits inside its own noise | the gate is void; measure the band, write a new gate |
+| `HELD_WITHOUT_MEASUREMENT` | a rule names an axis and nothing measured it | judge with the benchmark, or declare the run diagnostic — a gate that was checked is not a gate that was applied |
+| `NEEDS_SIGNATURE` | a band could not be verified, or a rule names no axis | a human decides; do not decide for them |
+| `UNRESOLVED` | the difference is smaller than the gate can resolve | this is **not** "no difference" — report the size needed to settle it |
+| `DIAGNOSTIC` | the run was declared diagnostic | no acceptance verdict may be drawn from it |
+
+Three of those exist because a flag can silence them. `allow_self_judged`,
+`allow_nonstandard` and `allow_missing` are meant to be deliberate acts, but
+specs get copied, so they ride on the verdict itself — a run comes back as
+`COMPLETE (declared: allow_self_judged)`, where a reader cannot skip past it.
+
+## The line
+
+Producing a candidate is mechanical, and every mechanical step here has been
+skipped by hand at least once. Each skip produced a number that looked real and
+was not.
+
+```text
+state purpose -> verify seal -> generate -> completeness gate
+             -> blind label / judge -> frozen gate -> persist -> status
+```
+
+The frozen gate is written **before** the run and the line may not change it. A
+loop that can move its own threshold optimises the threshold, not the work.
+What the gate reads matters as much as where it sits: conformance to a written
+portrait of the agent you want is a **diagnostic**, not a gate. A gate read
+from the portrait selects for resemblance; a gate read from the work selects
+for the work.
+
+Scoring is not done here. The judge stage calls the journeyman CLI, reads its
+report, and carries the axes into the run record. The dependency runs one way
+and never back — masterwork imports nothing from it, and a workshop without it
+gets a clear refusal instead of a surprise. One guard comes free with that
+contract: journeyman marks a run `self_judged` when the agent endpoint also
+served as the judge, and the line treats that as a gate rather than a
+footnote. A threshold applied to a self-judged score is the maker grading his
+own piece with extra steps.
+
+It costs nothing to start. A single local model is enough, because the gates
+are deterministic and the axes computed by replaying an agent's own events need
+no judge. In the run that set our bands, those free axes were also the
+steadiest — 0.02 across repeats, against 0.33 for the judged ones. A separate
+judge buys comparability with other people's numbers; it does not buy entry.
+
+A run leaves two kinds of thing behind, and the split is not "keep less".
+Transcripts, judge logs and per-cell records are local and disposable: they
+live under `runs/`, which version control ignores, and the seal names
+everything needed to re-derive them. The run record and the benchmark's report
+are archived on purpose — both are kilobytes, and between them they carry the
+seal, the grid, the axes, the gate and the verdict. That is what a later
+disagreement is settled with. `pipeline/retain.py archive <run-dir> <dest>`
+copies out exactly those two; oversized run directories are reported, never
+deleted, because deciding what to remove is not the line's call.
+
+## Making a piece
+
+<p align="center"><img src="https://raw.githubusercontent.com/codechu/masterwork/v0.0.1/assets/ceremony.jpg" alt="A figure writes alone by lantern light; every earlier work in the room is draped and turned away, and a stick of sealing wax lies unused on the table." width="520"></p>
+
+`pipeline/ceremony.py` runs the sitting. Four properties are the recipe rather
+than implementation detail, and each was paid for:
+
+- **Anchorless.** No earlier commitment or name is anywhere in the context.
+  With one present the model copies it, and a copy is a role worn rather than a
+  pattern re-derived — it also destroys the measurement, since you can no
+  longer tell whether the teachings carried.
+- **One session.** Every answer is an assistant turn in the same dialogue, so
+  at the end the model distils something it actually said, not a text handed to
+  it as its own.
+- **Order shuffled by a seed.** Answers otherwise echo their neighbours in the
+  order they were asked.
+- **The closing turn is not a rewrite.** It says: these are your words, remove
+  repetition, join where they meet, drop nothing. Asked instead to "write an
+  identity", a model produces a description of one.
+
+Teachings, questions and sampling profile are data. What lives in the code is
+the shape of the sitting.
+
+## Tales, not instructions
+
+What comes out is a character, not a role. The distinction is mechanical rather
+than philosophical:
+
+- **The corpus is third-person craft stories.** A story gives a pattern;
+  "behave like X" gives a role, and a role is worn and dropped. Assigned traits
+  do not survive a fresh context.
+- **What binds is the model's own sentence.** The identity file is not written
+  for the model; it is what the model wrote when asked. On every later request
+  it reads its own past words, not an instruction.
+- **The proof is unprompted behaviour.** Not "who are you" — a fresh,
+  same-class task that never mentions character. A role is performed when asked
+  for; a character shows up unasked.
+
+The piece that gets submitted is the sealed identity file: the model's own
+words, plus the maker's mark — corpus hash, script hash, question seed,
+sampling seed, date. Without those five it cannot be reproduced, so it cannot
+be sealed.
+
+## Writing your own corpus
+
+You get the line and its gates, the shape of the sitting, and the method
+written down: how a tale differs from an instruction, where a tale comes from,
+what a question has to do, why instructions come last, and how to tag a tale
+whose effect cannot be separated. That is
+[`corpus/GROWING.md`](https://github.com/codechu/masterwork/blob/v0.0.1/corpus/GROWING.md), and it is the part that transfers.
+
+You do not get a corpus. Three pairs come with the repository in
+[`corpus/starter/`](https://github.com/codechu/masterwork/blob/v0.0.1/corpus/starter/) so the machinery runs end to end and the
+form is unambiguous; they are a beginning and they are meant to be outgrown.
+The corpus is the work, and it has to be yours: a tale is written from
+something **you** measured and got wrong, and a corpus assembled from someone
+else's failures teaches your agent to avoid a life it never lived.
+
+So: take the line, take the rules, write your own tales, and expect the writing
+to be most of the work. If that sounds like too little to hand over, the honest
+answer is that the rest is not ours to give — it is the record of one house's
+mistakes, and yours will be different.
+
+## Failure modes
+
+A tool that produces the appearance of rigour is worse than no tool, and this
+one can. Two failure modes are made harder here; three are not automatable and
+belong to whoever is running it.
+
+Made harder: **declarations turning into defaults**, handled above by putting
+them on the verdict. And **a gap nobody measured** — the pair cutter takes
+`--gap-from`, the record of the judge's spread that justifies the threshold,
+and writes it into every pair. Without it you get a warning, because a training
+set built from a number someone chose is the quiet way to bake a judge's noise
+into weights.
+
+Not automatable, and this is the honest part:
+
+- **A gate can execute the wrong band.** The checker runs the command and
+  compares the result; it cannot know whether that command measures the thing
+  that actually varies. Choose a small band and your threshold looks
+  comfortable.
+- **PASS is not truth.** It means a threshold sat above a measured band.
+  Whether the axis is the one worth deciding on is a separate judgement, and a
+  wrong axis passes just as cleanly as a right one. We spent a day gating on
+  resemblance to a written portrait before noticing we should have been gating
+  on the work; no tool here would have caught that.
+- **A corpus can become a rulebook.** Nothing stops forty instruction-shaped
+  tales from being written. The result is a bloated prompt and the conclusion
+  that the method does not work, when what was tested was not the method.
+
+## Status
+
+Not published, and not proven. Our own corpus has eighteen tales and took
+months. It has produced eighteen candidates and none of them is ahead of the
+agent already running: the closest comparison came back `UNRESOLVED` rather
+than won — the difference was smaller than the gate could resolve, which is not
+the same as no difference.
+
+**What would change this:** one candidate separating from the incumbent through
+a gate frozen before the run, on an axis read from the work rather than from a
+portrait. That is also the condition for publishing. The name is honest about
+the position — a masterwork is submitted, and it can be refused.
+
+## Documentation
+
+| document | when you need it |
+|---|---|
+| [`AGENTS.md`](https://github.com/codechu/masterwork/blob/v0.0.1/AGENTS.md) | before running anything — the refusals, the vocabulary, and the four things not to do. Read it whether or not you are an agent |
+| [`docs/run-spec.md`](https://github.com/codechu/masterwork/blob/v0.0.1/docs/run-spec.md) | writing the JSON for one run, field by field |
+| [`docs/campaigns.md`](https://github.com/codechu/masterwork/blob/v0.0.1/docs/campaigns.md) | repeating an arm to measure a band, or comparing two arms against it |
+| [`docs/gates.md`](https://github.com/codechu/masterwork/blob/v0.0.1/docs/gates.md) | writing a threshold that can hold — start from [`gates/succession.md`](https://github.com/codechu/masterwork/blob/v0.0.1/gates/succession.md) |
+| [`corpus/GROWING.md`](https://github.com/codechu/masterwork/blob/v0.0.1/corpus/GROWING.md) | before writing a single tale — the shape is in [`corpus/starter/`](https://github.com/codechu/masterwork/blob/v0.0.1/corpus/starter/) |
+| [`CHANGELOG.md`](https://github.com/codechu/masterwork/blob/v0.0.1/CHANGELOG.md) | every entry names the incident that caused it |
+
+<details>
+<summary>Layout</summary>
+
+```text
+pipeline/   the line: one run, all gates, no manual steps
+tools/      instruments the line calls (blind labelling, gate checks)
+gates/      frozen gate templates — example.md is the smallest one
+examples/   runnable fixtures; the trace above comes from here
+corpus/     the method, and a starter of three tale-and-question pairs
+runs/       local run directories (ignored — heavy and re-derivable)
+```
+
+</details>
+
+---
+
+<p align="center"><img src="https://raw.githubusercontent.com/codechu/masterwork/v0.0.1/assets/icon.png" alt="Masterwork guild seal — a struck M hallmark inside a rope ring" width="96"></p>
+
+Contributing: [CONTRIBUTING.md](https://github.com/codechu/masterwork/blob/v0.0.1/CONTRIBUTING.md) · Changelog:
+[CHANGELOG.md](https://github.com/codechu/masterwork/blob/v0.0.1/CHANGELOG.md) · Licensed under the
+[Apache License 2.0](https://github.com/codechu/masterwork/blob/v0.0.1/LICENSE).