Your drafts, examined by someone else and ready to send with your query.
Agents are asking writers how their work was made. Your own word is exactly what is in question, and AI detectors settle nothing. So lodge the record your writing already left behind, have an independent examiner catalogue it and run defined procedures against it, then send agents a card of the findings with a note from you.
Lodge what you have
Upload the manuscript and the documentation behind it: version histories, revision snapshots, session logs, screenshots, notebooks, dated correspondence. Everything is hashed on receipt.
An examiner works through it
An independent examiner performs the same six published procedures on every submission and writes down what each one found, including what it could not show.
Send your locker card
Make one link per agent or publisher, each with its own note. They open a single tidy card: your photo and words on top, the registry's findings beneath.
What an agent sees
One card, two voicesYours to shape
Your photo, a short line about you, a note written for that agent or publisher, a link to your website. Pick a theme, an accent and a type style to make it yours.
Never restyled
The examination record, then one tidy entry for each kind of documentation: version histories, session logs, screenshots, notebooks, AI scans. Written by the examiner and set in fixed type, so it cannot be dressed up.
There is no score here, no grade, no pass mark and no seal of approval. A certificate would mean an agent had to trust our judgement about your work — and would mean writers whose process leaves few traces got marked down for how they write. Instead we report what you supplied and what defined procedures found in it. The reader draws their own conclusion, which is the only conclusion worth anything.
The six procedures, published in advance
Performed on every submission- Manuscript binding
Computes a SHA-256 hash of the manuscript file as submitted and records it on the published record.
What it cannot show: Nothing about how the manuscript was written. It fixes which file everything else in this record refers to.
- Catalogue integrity
Hashes every evidence item on receipt, opens each one, and confirms it is readable and of the type the author described.
What it cannot show: Whether the contents are genuine. It establishes what was received and that the examiner read those exact bytes.
- Coverage mapping
Plots the date range of every dated item against the composition period the author stated, and measures the days covered and the largest uncovered span.
What it cannot show: That covered days were spent writing, or that uncovered days were not. Coverage is a property of the documentation, not of the author.
- Internal reconciliation
Cross-checks independent items against each other: session-log arithmetic against draft word counts, revision timestamps against session times, stated revision counts against those present.
What it cannot show: That consistent records are true records. Consistency across independent sources is a fact worth reporting, not proof.
- Lineage sampling
Selects passages from the final manuscript at random, and searches earlier drafts and revision data for antecedent states of each one.
What it cannot show: That untraced passages were not written by the author. Absence of an antecedent in the supplied evidence is exactly that, and no more.
- Disclosure cross-check
Compares the author's written statement about the tools they used against what is observable in the catalogued artifacts.
What it cannot show: Tool use that leaves no artifact. A statement this procedure does not contradict has not thereby been confirmed.
How documentation is described
Every item is catalogued with a note on how it came to exist. That is a statement of fact, not a ranking — nothing scores better or worse, and no category disqualifies a submission.
Supply one if you have it and it will be catalogued as received, but no procedure is applied to it and it changes nothing. Detector output is a tool’s opinion about finished text, not a record of how that text was made. Those tools are unreliable in both directions and misfire on writers whose first language is not English. A record that leaned on one would deserve no trust.