Print the Bogen fainter, and let one word jump a frozen plan

[MarkusNeusinger]

Two owner wishes for the own-hand capture sheets (2026-09-06), both about what the FIRST printed Bogen has to be. No sheet has been printed yet, so neither change costs paper.

1 · A fainter lineature

Owner: "as long as it is faintly recognisable in the print, it is strong enough." Every ruling of the capture theme (core/eigenhand/geometry.py::CAPTURE_STYLES) moves down one step — the baseline takes the width the auxiliary lines had, and the auxiliary lines go thinner again:

Role	before	now
baseline	0.22 mm	0.12 mm
waist	0.15 mm	0.10 mm
ascender/descender (dashed)	0.12 mm	0.08 mm
slant grid (dashed)	0.10 mm	0.06 mm
word box + verdict box	0.12 mm	0.12 mm (unchanged)

Why thinner is the safe direction: the lineature only has to guide the hand for the seconds a row takes, and everything it does afterwards is damage. Faint cyan still prints — 0.06 mm is 1.4 dots at 600 dpi, so the device draws its thinnest line rather than nothing — while every millimetre of cyan that is NOT on the paper is one the ingest cannot mistake for ink, cannot leave in the blue channel and cannot smear into a stroke the fit measures. The box role keeps 0.12 on purpose: the same role draws the verdict box in the right margin, which the writer has to find and tick (a forgotten tick discards the row).

The order baseline > waist > ascender/descender > slant survives, so the row still reads at a glance. Positions, Passmarken and the printed ruler check are byte-identical — only stroke widths moved, which is why every layout test and every cfg stamp stayed put and only the golden PDF hash is re-pinned (declared re-baseline, with the reason next to it).

The owner's print task text stays valid as written: colour printer, A4, 100 %, ruler on the Passmarken at 190.0 / 277.0 mm. data/samples/own-hand/README.md gains one rule saying the lineature is deliberately this faint — a print in which it is barely visible is the right one, not a printer fault.

2 · „Kurrentschrift" on the first sheet — the Anheftung

The word is the site's hero and the author wants a written specimen early. It was not in the Wortvorrat, and the strip plan is frozen at 180 strips across three waves, so it could not be put INTO a strip: append-never (Prüfstein 4) forbids exactly that.

The mechanism is the smallest one that fits the doctrine instead of bending it:

• tools/eigenhand/corpus.py gains a pin curation layer with PINNED_FIRST — words that earn their place by what they are, not by their coverage. So the word is a normal pool word: it enters the Soll universe and is counted in Bestand and coverage like any other.
• python -m tools.eigenhand.pool pin appends ONE strip of its own per pinned word and registers it in the plan's new pins block. verify_immutable runs on this path too; the diff to streifen.json is purely additive.
• core/eigenhand/plan.py::ordered_strips puts pins at the head of plan order. The pin is expressed in plan order rather than in the queue so that every reader agrees on what comes first — print queue, coverage progression and Bestand walk the one order. A pin leaves the front the ordinary way: once the strip is belegt it is out of the queue.

Result: S0181 = „Kurrentschrift", row 1 of the next Bogen (B0001, seven rows, then S0001 … S0006 as before). Spelling: the plain word, which core/shaping.py already shapes as Kurrentſchrift — the default rules give the long ſ at the start of the second morpheme -schrift, so no fugen form is needed and the printed label carries no hint marks. pins is additive, so the plan format stays 2; a second pool pin run is a no-op ("already planned, not pinned again").

Three decisions taken here and written down in the proposal: the plan stays hand-agnostic (one plan serves all three scripts — a hand-scoped pin would need hand-scoped plans), a word already planned is never pinned again, and a pinned strip is not filled up with quota words (that would need the local-only Übergangsraum weights; the box generator hands the free row width to the one word, so „Kurrentschrift" gets the full 180 mm).

Verification

• /verify-core: full suite green (2474 passed, 18 skipped), ruff check + ruff format --check clean.
• Direct invocation instead of the synthetic smoke, because the diff is the sheet: one Bogen composed and rendered to PDF before/after, rasterised at 300 dpi, plus a 4× crop of the same row in both versions. The row reads clearly and prints faintly.
• tools.docs_budget check, tools.docs_register check, tools.changelog check green.
• No bench numbers: this core/ change touches page geometry and plan order, not extraction, composition or glyph rendering.

New tests: the capture lineature stays strictly fainter than the app's druck theme, keeps its hierarchy and stays printable; a pin appends and leads without touching a frozen strip; an already-planned word is not pinned again; an uncurated word is refused; the committed plan pins what the curation pins; the pinned strip is row 1 of the first sheet, and a written pin leaves the front of the queue.

New term Anheftung is in docs/reference/glossar.md (§5 + Schnellindex). It is not in kurzglossar.md: that file's own Nachzieh rule (docs/dokument-status.md) is the two-of-three threshold, and the term reaches one source so far.

🤖 Generated with Claude Code

https://claude.ai/code/session_01UEScQMZFvxxNNyNJYryfa3

[Copilot]

🟡 Changes recommended

The repeatable pin path can combine multiple pinned words into one strip, violating the documented one-strip-per-pin contract.

Once you've addressed the issues Copilot identified, you can request another Copilot review.

Pull request overview

Refines Eigenhand capture sheets and adds pinned words to the frozen strip plan.

Changes:

• Makes capture rulings fainter while preserving layout geometry.
• Adds append-only pinned strips, led by “Kurrentschrift”.
• Updates tests, documentation, plan data, and changelog.

File summaries

File	Description
tools/eigenhand/pool.py	Adds pinned-strip generation and CLI support.
tools/eigenhand/corpus.py	Curates the first pinned word.
core/eigenhand/plan.py	Places pinned strips first.
core/eigenhand/geometry.py	Reduces capture ruling widths.
core/eigenhand/streifen.json	Appends pinned strip S0181.
tests/test_eigenhand_stack.py	Tests print-queue pin behavior.
tests/test_eigenhand_pdf.py	Re-baselines PDF output.
tests/test_eigenhand_geometry.py	Tests capture ruling properties.
tests/test_eigenhand_coverage.py	Tests pin creation and immutability.
docs/proposals/eigenhand-erfassung.md	Documents both design changes.
docs/reference/werkzeuge.md	Documents the pin command.
docs/reference/glossar.md	Defines “Anheftung”.
data/samples/own-hand/README.md	Updates printing instructions.
changelog.d/eigenhand-feine-lineatur-und-anheftung.md	Records user-visible changes.

Review details

Suppressed comments (1)

tests/test_eigenhand_geometry.py:77

• The numeric explanation is reversed: one 600 dpi dot is about 0.042 mm, so 0.05 mm is slightly more than one dot, not under one. Correcting this keeps the test's physical threshold understandable.

        # A line the printer cannot resolve is guidance that never reaches the
        # paper: 0.05 mm is under one dot at 600 dpi (0.042 mm).

• Files reviewed: 14/14 changed files
• Comments generated: 2
• Review effort level: Balanced

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

[MarkusNeusinger]

Both review findings are fixed in 743f475.

• tools/eigenhand/pool.py: every fresh pinned word now gets a strip of its own instead of going through the wave packer. The docstring, the proposal, the glossary and the README all promised exactly that — one strip, full row width — and pins are few and deliberate, so keeping the promise is better than saving a row. A test with two short pins ("das", "lesen") now covers the case the packer would have merged; the committed plan is unchanged, since it holds a single pin.
• tests/test_eigenhand_geometry.py: the hierarchy assertion compares adjacent widths with > instead of against a reverse-sorted copy, so two roles at the same width fail as they should. The printable floor's comment also had its arithmetic backwards (0.05 mm is about one dot at 600 dpi, not under one) and now says what the floor means.