Changes in version 1.0.1 (2026-09-03) CRAN preparation release - Added complete source attribution and copyright notices for bundled source-derived instrument content. - Made the software citation derive its version and authors from package metadata and completed the author list for the 2026 updated-CFIR article. - Added an installed HTML technical guide alongside the PDF guide. - Added a dedicated R 4.1 continuous-integration check for the documented minimum R version. - Updated release metadata and submission checks for an initial CRAN review. - Corrected the complete five-category partition by adding n_neutral_complete and deriving all five category counts from pcat_class5, while retaining directional n_neutral. - Extended small-cell suppression to derived modal classifications. - Required valid direction and effect parsing before assigning a non-invalid strength category. - Added validation of every page before failure-safe multi-page profile PDF writing, so a late page error cannot leave or replace a partial target. - Enforced true finite scalar validation for consensus and action-plan probability thresholds and repaired broken links in the source archive. - Added pct_complete_class, the share of valid-direction responses with a complete five-category classification, as an explicit completeness diagnostic. - Rejected directory-valued profile PDF targets and infinite page dimensions before any output or backup operation. - Represented polarization as unavailable rather than false when the required directional percentages are missing. - Revised examples following CRAN review: short executable examples now run normally, interactive Shiny use is guarded, optional dependencies are checked, and demonstrated file output uses temporary locations with cleanup. - Corrected DESCRIPTION quotation so the pCAT and CFIR acronyms are unquoted. - Updated package authorship and software-copyright metadata to list Lilac Li as package author and joint copyright holder while retaining Jae Man Park as package maintainer and joint copyright holder without the package-author role. Changes in version 1.0.0 Stable GitHub release - Added pcat_analyse() and a structured pcat_analysis result object. - Added pcat_write_analysis() for reproducible export of validation issues, classified responses, item summaries, consensus diagnostics, action plans, an analysis manifest, and an optional multi-page profile PDF. - Rebuilt visualization with ggplot2, shared legends, readable facet labels, explicit neutral labels and denominators, and device-independent output. - Improved profile defaults for multi-panel readability and expanded absent item records explicitly in respondent and change heatmaps. - Added pcat_save_profile_pdf() to place one grouping combination per page. - Added original and updated CFIR mapping metadata, including item 10's secondary Funding mapping from the 2026 supplementary table. - Added user-defined small-cell suppression in pcat_summarise(). - Clarified denominators: directional percentages use valid direction responses; five-category strength percentages use complete direction/effect responses. - Added a packaged technical user guide accessible through pcat_user_guide(). - Added GitHub Actions, pkgdown configuration, citation metadata, issue and pull request templates, security/support/governance policies, and release tools. - Added automated tests and pcat_self_test(). - Added privacy-aware export controls, analysis settings/session provenance, and deterministic cleanup of stale package-generated output files. - Preserved the methodological boundary that item display codes must not be summed into a purported validated total pCAT score. Changes in version 0.1.2 - Corrected restoration of graphics parameters in early base-graphics plots. Changes in version 0.1.1 - Corrected early development-test and installation issues. Changes in version 0.1.0 - Initial development release.