Skip to content

v0.6.2

Choose a tag to compare

View all tags
@wfukatsu wfukatsu released this 11 Jun 02:16
· 361 commits to main since this release
v0.6.2
0484e03

What's New

Agent Component Audit Fixes

A full audit of the agent-constituting components (skills, hooks, rules, manifests) resolved the following:

  • Hooks now block correctly: validate-frontmatter.sh and validate-mermaid.sh emit errors to stderr and exit 2 in hook mode, so Claude Code feeds validation failures back to the model for self-correction (exit 1 was user-visible only). CLI mode keeps exit 1
  • Hooks registered with the plugin: new hooks/hooks.json registers both validators on PostToolUse (Write|Edit|MultiEdit) — they now fire automatically on plugin install
  • Mermaid bracket check fixed: quoted label content is stripped before bracket-balance counting, eliminating false positives on labels like A["決済 (カード)"] and restoring detection of genuinely unbalanced brackets
  • Stale reference paths fixed: 10 scalardb skills pointed at the legacy .claude/docs/ and .claude/rules/ layout; all now use ${CLAUDE_PLUGIN_ROOT}/skills/common/references/ and ${CLAUDE_PLUGIN_ROOT}/rules/
  • All rules now reachable: previously unreferenced rules (exception handling, CRUD/JDBC/2PC patterns, schema design, config validation, Java best practices) are explicitly referenced from their relevant skills
  • English-only SKILL.md enforced: Japanese template text removed from review-report and report skills; localization is driven by output_language
  • Manifest fixes: review-report skill registered in marketplace.json; plugin versions aligned with release tags (0.5.0 → 0.6.2)
  • Reduced session context: CLAUDE.md no longer auto-imports the full ScalarDB rule set (~20k tokens) into every session; rules are read on demand
  • Consistency fixes: ddd-tactical-architecture-evaluation.md filename aligned across templates and dependency manifest; model: field added to 10 nested migration skills; lowercase skill.md normalized to SKILL.md

Previous Changes (included since v0.6.1)

  • New create-domain-story skill: Domain Storytelling visualization per bounded context (interactive 7-stage facilitation or auto-generation), integrated into the pipeline and HTML report
  • New review-report skill: quality review of the generated HTML report (completeness, score accuracy, Mermaid syntax, language consistency, structural integrity)
  • Added ec-monolith sample project for end-to-end pipeline validation
  • Mermaid validator multi-line block parsing repaired (macOS awk compatibility) and ubiquitous-language term alignment rule added
  • Codex compatibility gaps in migration skills resolved; investigate skill gained calculation procedures and self-verification
Assets 2
Loading