v4.0.0 — Thirty-two project skills + craft-grade bilingual site

Major release. The project skill library grows from twenty to thirty-two skills, completing operational coverage of the research lifecycle, and ships together with a craft-grade bilingual overhaul of the companion site copy.

Added

• Twelve new project skills under .claude/skills/, each authored to the repository's bilingual skill schema: prisma-scoping-review-pipeline, sensitive-data-anonymization-gate, open-science-release-packager, authorship-contribution-ledger, peer-review-confidentiality-protocol, multilingual-concept-validity-audit, grant-proposal-workpackage-builder, teaching-feedback-ai-boundaries, public-scholarship-ethics-adapter, preregistration-analysis-plan-ledger, agent-portability-matrix, cross-agent-second-opinion. Each operationalizes the booklet it pairs with.

Changed

• The existing twenty skills were enriched in place, additively, with when-not-to-use boundaries, named cross-skill handoffs, a session-end record line, and a closing quality gate, without altering their proven eight-section format.
• The companion site copy (web/index.html) was rewritten for craft in both languages, the status band reinstated (thirty-three booklets, fourteen categories, thirty-two skills), and the skill catalogue advanced to thirty-two across the README pair, the catalog, the validator, and the plugin and marketplace manifests.

Cite via the Zenodo concept DOI 10.5281/zenodo.20289687, which resolves to the latest version.

v3.2.0

Twelve continuation booklets expanded bilingually from the author's source drafts, opening two new categories (013 Teaching and Supervision, 014 Tool Portability) and extending six existing ones. Released catalog: 21 to 33 booklets across 14 categories.

All 73 cited DOIs verified live against Crossref and doi.org; two mis-keyed identifiers corrected (Resnik and Elmore in 010-03; Yildiz ISBN in 007-05). Zero fabricated. Aggregate verified declarations 354 to 566.

Reading order rewired; 012 no longer the closing booklet. Navigation, catalog, landing index, AI-disclosure aggregate, READMEs, and roadmap updated.

v3.1.0 — bilingual content overhaul + journal redesign

A full bilingual content overhaul of all 21 booklets, shipped together with the academic-journal site redesign.

Content overhaul

• All 21 booklets rebuilt from the author's hand-revised Turkish sources. Turkish is canonical and verbatim; English is natively re-authored (not literal translation), bilingual parity preserved.
• Every reference re-audited against Crossref and doi.org with an independent refute pass. 18 mechanical fixes. 2 fabricated grey-literature citations in 003-01-0001 replaced with verified real documents (EDPB Opinion 28/2024; the 2025 KVKK special-category personal data guide); STM 2025 title corrected. Zero fabricated citations remain.
• Booklet 004 gains 3 archive/MOC diagrams with bilingual alt text. Dropped body tables restored to 7 booklets in both languages. Titles reconciled to the author's canonical forms (e.g. 001-01-0001 "Sosyal Bilimcinin Gözüyle Claude Code"; vault rendered as arşiv).

Redesign

• Landing page and MkDocs booklet theme redesigned in an academic-journal identity (teal + gold, self-hosted variable fonts, restrained motion). Zero third-party font requests.

Verification

• validate + lint (0 errors) + mkdocs build green; check-dois 87 alive, 0 dead.
• Stats unchanged: 21 booklets, 354 verified citations, 0 fabricated.

Full changelog: see CHANGELOG.md [3.1.0].

v3.0.2 — Provenance and licensing accuracy

Patch release. Corrects the published copyright record and strengthens provenance with a verifiable, registration-free protection stack.

Licensing accuracy. The unfiled U.S. Copyright Office (eCO) claim is removed from the LICENSE and the project surfaces. The copyright basis is restated accurately as automatic protection under the Berne Convention (Türkiye, Greece, Ireland, 180+ member states), evidenced by Zenodo, OpenTimestamps, and the public Git history.

New provenance layer. provenance/timestamp.txt + timestamp.txt.ots: an OpenTimestamps proof anchoring the release to the Bitcoin blockchain.

Per-booklet license marking. All 42 booklet footers now carry a CC BY-NC-SA 4.0 line.

Hygiene. Stale version labels fixed and made durable; version cascade 3.0.1 → 3.0.2 across all manifests.

No booklet content, citation, or count changed: 354 verified, 0 fabricated, 21 booklets, 20 skills.

v3.0.1 — zero-defect audit remediation

Patch release. A repository-wide zero-defect audit (63 Opus agents across all 21 booklets, every flagged citation re-verified live against Crossref) found and corrected 69 confirmed defects in published content. The reference set and aggregate counts are unchanged: 354 verified declarations, 0 fabricated.

Highlights: Hosseini et al. (2023) bibliographic fields corrected across 8 booklets; Munafò, Bramer, Xu, Wicherts, Resnik article-numbers/issues fixed; Schulhoff 31-author APA form; Valmeekam cited to NeurIPS proceedings; 24 Turkish prose semicolons split; country name standardized to Türkiye in English booklets; the author writing-style technical register applied (de-calque term, em-dashes removed from English prose); bilingual parity restored; footer word counts recomputed.

See CHANGELOG.md [3.0.1] for the full list.

v3.0.0 — The masterpiece milestone: full grid, doubled skills, review-ready JOSE paper

Major release closing the v3.0 arc. Twenty-one bilingual booklets with at least one in every one of the twelve categories, twenty companion skills with machine-enforced bilingual parity, 354 verified citation declarations with zero fabricated, and a Journal of Open Source Education paper refreshed to this finished surface.

Added

• A fifth distinctive contribution in the paper: the anti-AI-trace revision method, grounded in DOI-verified evidence that detection tools are unreliable (Weber-Wulff et al., 2023) and biased against non-native English writers (Liang et al., 2023), with the disclosure-over-detectability position stated explicitly.
• meta/jose-submission.md, the JOSE readiness checklist. The dual-license question is documented as an open maintainer decision; the submission itself is maintainer-gated and not performed by this release.
• A draft-pdf.yml workflow compiling the paper with the official Open Journals action.

Changed

• paper/paper.md moved from the v2.2.0 narrative to the v3.0.0 surface, with the quality-control section matched to current CI (validator with skill schema, markdownlint, pytest on 3.9/3.12, cff-validator, gitleaks).

Unchanged

• No booklet, citation, or skill content changed. Aggregate metrics: 354 verified declarations, 0 fabricated, 21 booklets at release status.

Full notes in CHANGELOG.md.

v2.9.0 — Booklet wave B: all twelve categories complete

Four new booklets with live-verified citation cores raise the released total from seventeen to twenty-one. Every one of the twelve categories now carries at least one released booklet, completing the catalog grid the v3.0 milestone builds on. Aggregate verified citation declarations rise from 306 to 354, with zero fabricated.

Added

• 003-03-0001 Material Passport: Tracking Sources Across Sessions — a six-field identity line for every source, FAIR-aligned at the personal archive layer, with the rule that a source without a passport does not enter the bibliography (6 verified refs).
• 005-02-0001 Ritual Hooks: Daily Logging, Session Persistence, Idle Time — research discipline moved from willpower to infrastructure, with guard hooks for secret scanning and citation protection (6 verified refs).
• 006-01-0001 MCP for the Researcher: What, Why, When — looking instead of remembering, a trust triage for third-party servers, least-privilege setup, and an honest section on when no bridge is needed (6 verified refs).
• 011-01-0001 Slides, Posters, and Lightning Talks with AI Assistance — three compressions of one study, visual integrity rules, and disclosure awareness on stage (6 verified refs).

Citation discipline
Nine new candidate DOIs verified live against Crossref before drafting. One candidate (Lally et al., 2010) was dropped over ambiguous registered metadata rather than cited with an uncertain date.

Infrastructure
The Pages workflow now runs on the Node 24 actions runtime, opted in ahead of GitHub's 2026-06-16 switch and proven on this release's own deploy.

Full notes in CHANGELOG.md.

v2.8.0 — Booklet wave A: four booklets with live-verified citation cores

Four new booklets join the catalog, raising the released total from thirteen to seventeen and aggregate verified citation declarations from 248 to 306, with zero fabricated. The academic writing (007), data analysis (008), and peer review (010) categories now have every planned slot at release status.

Added

• 007-01-0001 IMRAD Scaffolding: A Bilingual Approach — Turkish-first drafting, English re-authoring, parity discipline for titles, abstracts, and keywords, grounded in the documented costs of publishing in a second language (8 verified refs).
• 007-03-0001 Journal Fit and Cover Letters — fit as a researchable question, index layers including ULAKBIM TR Dizin, predatory journal screening, the cover letter as a document of fit (7 verified refs).
• 008-03-0001 Qualitative Coding with AI Assistance and Human Oversight — codebook design, the model as a second coder under protocol, reliability checkpoints, verbatim quote integrity, anonymization before model contact (6 verified refs).
• 010-02-0001 Anti-AI-Trace Writing for Revisions — the theoretical ground of the anti-ai-trace-revision skill: detection tools fail measurably and unevenly, integrity lives in disclosure, and the two-layer revision method this guide uses on its own prose (8 verified refs).

Citation discipline
Every DOI verified live against Crossref before drafting; quantitative claims checked against publisher or PubMed Central full texts. One candidate without a registered DOI (Sollaci & Pereira, 2004) was dropped rather than cited unverified.

Full notes in CHANGELOG.md.

v2.7.0 — Twenty project skills with bilingual usage notes

Minor release. The companion project skill set grows from ten to twenty, covering the research lifecycle from literature scoping through bilingual drafting, analysis discipline, citation verification, and AI disclosure to review response and release integrity. Every skill now carries a Turkish usage section alongside its English protocol, extending the guide's bilingual parity principle to the skill layer.

Ten new skills

anti-ai-trace-revision, bilingual-manuscript-scaffold, journal-fit-screening, qualitative-coding-discipline, statistical-consultation-protocol, research-ritual-hooks, research-lifecycle-pipeline, mcp-research-stack-triage, source-passport-ledger, conference-materials-bilingual.

Infrastructure

social-cc doctor environment check, a Python test job in CI on 3.9 and 3.12, a version single-source fix, a skill template, and a skill proposal issue template.

Integrity

No booklet content, citation, or DOI changed. Aggregate metrics unchanged at 248 verified declarations, 0 fabricated, 13 booklets at release status.

The Zenodo version DOI for v2.7.0 is minted on this release and recorded in a follow-up commit. The concept DOI 10.5281/zenodo.20289687 always resolves to the latest version.

v2.6.0 — Bilingual de-AI and naturalness pass

A bilingual language and naturalness pass over the companion website and all thirteen released booklets, in both Turkish and English.

• Removed the machine-writing rhetorical architecture (number-announce headings, mechanical rule-of-three lists, repeated antithesis tails) and remaining concept-level calques.
• Turkish Vault metaphor rendered as arşiv (the X-olarak-Y calque broken); parite → eşlik. English keeps Vault as the Obsidian term.
• Content, citations, frontmatter keys, heading counts, and bilingual parity preserved. Aggregate metrics unchanged: 248 verified declarations, 0 fabricated, 13 booklets at release.

Concept DOI (resolves to latest): 10.5281/zenodo.20289687