v1.8.0

Added — sprint v1.8 linter v2 + DRY + DX + index

• Linter v2 (writer/scripts/lint.py): five new categories —
  MARKETING_HYPE (revolutionary / game-changing / world-class / industry-leading /
  cutting-edge / best-in-class / groundbreaking / next-generation / state-of-the-art /
  unparalleled / unmatched + RU equivalents); EMPTY_CTA (click here / tap here /
  learn more / get started without object + RU); WEAK_OPENER (we're excited to /
  thrilled to / proud to + RU); VAGUE_BENEFIT (save time / boost productivity /
  get more done + RU); WRONG_TENSE_RELEASE (will support/enable/etc. in
  release-notes context, severity nit). Total: 28 categories.

• Code-fence skip in linter: lines inside fenced ``` / ~~~ blocks are no longer
  scanned. Previously skill_skills-update_input had 74 lines of false-positive
  hits matching marketing words inside JSON examples; now clean.

• Severity tags (blocker / caution / nit) per hit + aggregate
  by_severity in JSON output. Nit-level hits don't escalate the overall
  verdict. New CLI flags: --scan-code-blocks (opt-in for calibration files),
  --quiet (used by make lint-all).

• 7 new before-after.md calibration files: canon-check, essay-write,
  skills-update, style-check, translation-sync, viral-text, cold-email.
  Every skill (17/17) now has a calibration file showing the concrete
  transformation it produces.

• 3 expanded walkthroughs: viral-post.md (129→234), tone-shift.md (162→276),
  image-prompt-cover.md (165→351). Added RU port scenarios, A/B variants,
  multi-stage shifts, cross-model comparisons, refinement decision trees,
  and per-walkthrough troubleshooting sections.

• Skill tags in skills.json: closed-dictionary tags: [] array per skill
  (domain: fiction / non-fiction / marketing / social / product / tech-docs /
  ux-copy / visual / outreach; function: editing / generation / audit /
  translation / ops). scripts/validate.sh validates the dictionary.

• docs/SKILL-INDEX.md — auto-generated by scripts/gen-skill-index.py
  (make gen-index). Groups all 17 skills by layer, by domain, and by
  language. scripts/check-docs-consistency.sh gains a 6th sub-check
  verifying SKILL-INDEX freshness.

• common/references/ — shared cross-skill anti-pattern catalogues:
  banned-patterns-hype.md, banned-patterns-preambles.md,
  banned-patterns-empty-cta.md. install.sh gains install_shared_refs()
  to copy them to $PREFIX/common/references/. cold-email,
  landing-copy, and release-notes banned-patterns.md now cross-link
  to the shared files instead of duplicating their content.

• DX: make lint-all (writer linter across every reference, example, and doc
  file — advisory); scripts/install-precommit-hook.sh (installs local
  .git/hooks/pre-commit that runs the linter on staged .md + smoke gate, skips
  calibration files); .github/PULL_REQUEST_TEMPLATE.md checklist extended to
  all 5 CI gates + tag-dictionary + gen-readme/gen-index + common/references
  consideration.

Changed

• writer description: "23 categories" → "28 categories" (reflects linter v2).
• viral-text, release-notes, rfc-writer SKILL.md descriptions: explicit
  "Wraps writer" mention (consistency with other wrappers — moved from v1.7
  prep into shipped commit on this branch).
• cold-email/references/banned-patterns.md, landing-copy/references/banned-patterns.md,
  release-notes/references/banned-patterns.md: lead with cross-link to
  common/references/ shared catalogues.
• scripts/smoke.sh: writer linter self-test now passes --scan-code-blocks
  so calibration BEFORE samples inside ``` fences still trigger neuroslop verdict.