Part of netresearch/skill-repo-skill#157; ticks the `german-technical-writing` checkbox. ## What was cut `skills/german-technical-writing/references/register.md`: 1202 -> 964 words (-19.8%, re-measured on `origin/main` at HEAD `b6401e1`, not the weeks-old audit baseline). Cut per the [content value rubric](https://github.com/netresearch/skill-repo-skill/blob/main/skills/skill-repo/references/skill-quality.md) (deletion test: remove it, is a fact lost?): - **Duplicate scope note** (`### Commit messages, MR/PR descriptions, and release notes — not in scope`, 3 sentences): the same fact already lives in this file's own intro line and in `SKILL.md`'s Scope note — a third full restatement adds no fact, just words. - **Bare "avoid mixing active and passive" rule**: no example, no anglicism tie-in, generic writing advice applicable to any language — unlike every other Voice/Tense/Person rule in the file, which carries a ✅/❌ pair. - **Redundant "Style" bullets** under Jira Issue Description: 2 of 3 bullets ("impersonal", "present tense") repeated the Tense/Person sections verbatim in the same file; kept the one bullet with no earlier statement (parallel bullet form) and pointed back to the sections above instead of re-asserting the rule. - **`## Orthography Minimums`** (ß/ä/ö/ü, no-ASCII-folding, basic capitalization): baseline German competence any writer (or model) already applies without a reference file — not the anglicism/register differentiator this skill exists for. Regenerable from a one-line prompt. - **Internal German Wiki / Spec Pages subsection**: 4 bullets, 3 of which were verbatim restatements of the Person section's Sie/Du rule or generic Markdown heading-hierarchy advice unrelated to German register. Kept the one bullet with new information (specification-section vs. how-to-section voice split) and pointed the Sie/Du choice back to the Person section. - `README.md`: dropped the "ß/ä/ö/ü orthography" clause from the Register-enforcement feature bullet — it described the now-removed Orthography Minimums section and would otherwise be a dangling claim. ## What was kept (and why) - **Every present-indicative / impersonal-voice rule** (Tense, Voice, Person sections) — unchanged, including every ✅/❌ example pair. - **Every concrete example** across Tense, Voice, Person, Compound Nouns. - **Jira Issue Description template** (the `h2. Problem/Ziel/Akzeptanzkriterien/...` code fence) — untouched; evals #2 and #11 assert on these exact headings. - **Jira Comment conventions** (h3 headings, table syntax, `{{monospace}}`, hedging phrases to cut) — concrete, artifact-specific, not generic. - **Sentence Length numeric thresholds** (10-15 / 25 / 35+ words) and the reasoning tying long sentences to anglicism risk — a non-obvious insight specific to this skill's mission, not generic writing advice. - **Compound Nouns section** in full — concrete escalating examples plus the hyphenation rule. - The anglicism-substitution lexicon (brechen->werfen, gefangen->abfangen, returnen->zurückgeben, triggern->auslösen, failen->fehlschlagen, die Test->der Test) and the Denglisch-loanword lexicon (der Commit, die Pipeline, der Branch, die Exception) — **not in this file at all**; they live in `references/anti-patterns.md` and `references/lexicon.md`, untouched by this PR. ## Kept, borderline - **Checklist item 6** ("Are ß/ä/ö/ü intact?") — left in place even though its backing `## Orthography Minimums` section is gone. It's a self-contained one-line reminder, not a link/pointer, and stays meaningful without the explanatory section. - **Sentence Length "reason" paragraph** ("Jira readers skim. Long Verbalketten-Sätze are where anglicisms hide...") — borderline generic-sounding, kept because it draws an explicit, non-obvious causal link between sentence length and anglicism risk (the skill's core concern), not a restatement of common writing advice. - **Jira Issue Description code-fence template's bracketed instructional comments** (e.g. `[Impersonal, present tense, 2-4 sentences max.]`) — left fully intact rather than compressed; this is the eval-tested, load-bearing template, not padding. ## Re-verification against the (weeks-old) audit The audit target's "~25%" estimate was re-checked against the current `origin/main` tip (`b6401e1`, includes the unrelated `#12`-`#15` register.md/SKILL.md commits merged since the audit ran — register.md gained one checklist item in the meantime, unrelated to this cut). After preserving every hard-protected category (present-indicative/impersonal-voice rules, every concrete example, org specifics, the eval-tested template), the genuinely generic/duplicate content measured out to **19.8%**, not 25% — the gap is content the audit likely also flagged but that turned out to carry real information once checked line-by-line (e.g. the wiki voice-split nuance, the Sentence Length reasoning). Reported as measured, not adjusted to hit the estimate. ## Verification ``` $ wc -w skills/german-technical-writing/references/register.md 964 skills/german-technical-writing/references/register.md # was 1202 on origin/main $ bash <skill-repo-skill>/skills/skill-repo/scripts/validate-skill.sh . ... OK: SKILL.md is 499 words (under 500 limit) Errors: 0 Warnings: 6 # pre-existing README section warnings, unrelated to this change $ bash <skill-repo-skill>/scripts/audit-skills.sh skills REFERENCES: 5 total -> 5/P1 + 0/P2 + 0/P3 = 5 reachable, 0 ORPHAN ``` `npx markdownlint-cli2 skills/german-technical-writing/references/register.md README.md` — 0 errors. `pre-commit run --files README.md skills/german-technical-writing/references/register.md` — all applicable hooks passed. **Pre-existing, out of scope for this PR:** `audit-skills.sh` reports SKILL.md's `description` field at 1985 chars (FAIL, cap 1536) — introduced by the unrelated "Two opposite traps" section merged in `#12`/`#13`, before this branch was cut. Not touched here; flagging for a separate PR. Not merging per task instructions — ready for review.

Uh oh!

ChatGPT reports invalid SKILL.md frontmatter on upload #2

Description

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions