chore(release): bump version to v0.5.20

chore(release): bump version to v0.5.19

chore(release): cut v0.5.17 with commit-staged guardrails

- templates/commands/commit-staged.md: remove Session Context Awareness
  section and add explicit "staged diff is the only source of truth"
  guardrail to prevent hallucinated intent from prior conversation turns
- CLAUDE.md: sync /codexspec:commit-staged description and add a
  Self-Bootstrap rule forbidding manual edits to .claude/commands/codexspec/
  for distributed commands (treat as install artifact)
- pyproject.toml, uv.lock: bump version 0.5.16 -> 0.5.17

fix(docs,i18n): migrate glossary path to canonical docs/i18n location

- Update glossary path references from .codexspec/i18n to docs/i18n/
- Fix internal maintenance command templates (check-i18n-semantics, translate-docs)
- Update CI workflow documentation
- Correct Enhanced Commands count in CLAUDE.md (6 → 4)
- Update translation files and related tests
- Synchronize all affected templates and configurations

 feat(review): add scoring rubrics, score justification, and challeng…

…e response protocol (#2)

* feat(review): add scoring rubrics, score justification, and challenge response protocol

Add transparent scoring system to all 5 review templates and quality targets to 2 generation templates.

Review templates (review-spec, review-plan, review-tasks, review-python-code, review-react-code):
- Scoring Rubrics with score ranges and typical deductions per category
- Enhanced Scoring Breakdown with Rubric Basis and Deduction Details columns
- Suggestion Score Cap limiting suggestions to max 5 points impact
- Score Validation Checklist for arithmetic and consistency verification
- Score Challenge Response Protocol for objective 3-step dispute handling

Generation templates (spec-to-plan, plan-to-tasks):
- Quality Targets section aligned with corresponding review rubric categories

* chore: update marketplace.json to v0.5.15

feat(review): add scoring rubrics, score justification, and challenge…

… response protocol

Add transparent scoring system to all 5 review templates and quality targets to 2 generation templates.

Review templates (review-spec, review-plan, review-tasks, review-python-code, review-react-code):
- Scoring Rubrics with score ranges and typical deductions per category
- Enhanced Scoring Breakdown with Rubric Basis and Deduction Details columns
- Suggestion Score Cap limiting suggestions to max 5 points impact
- Score Validation Checklist for arithmetic and consistency verification
- Score Challenge Response Protocol for objective 3-step dispute handling

Generation templates (spec-to-plan, plan-to-tasks):
- Quality Targets section aligned with corresponding review rubric categories

feat(commands): add quick command for one-stop automated SDD workflow

- Create /codexspec:quick template for streamlined requirement-to-implementation flow
- Consolidates spec generation, planning, task breakdown, and code implementation
- Supports complexity assessment, concise clarification, and auto-review at each stage
- Add translation entries for all 8 languages (en, zh-CN, ja, ko, es, fr, de, pt-BR)

fix(templates): use Skill tool to invoke review commands instead of d…

…uplicating logic

- generate-spec: call /codexspec:review-spec via Skill tool
- spec-to-plan: call /codexspec:review-plan via Skill tool
- plan-to-tasks: call /codexspec:review-tasks via Skill tool

This ensures review commands are properly invoked rather than having
generate commands directly write review reports.

feat(i18n): add commit language consistency check to init command

When users run 'codexspec init --lang <lang>' on an existing project,
the output language gets updated but the commit language may remain
different. This change adds a prompt to ask users if they want to
update the commit language as well.

- Add get_commit_language() to extract commit language from config
- Add update_output_language() to update only output language
- Add translations for 8 languages (en, zh-CN, ja, ko, es, fr, de, pt-BR)
- Prompt user when commit language differs from selected language

feat(commit-staged): add safety mechanisms to prevent staging area mo…

…dification

- Add explicit forbidden-tools in frontmatter (git add/reset/checkout/restore/stash/rm)
- Add "Forbidden Operations (CRITICAL)" section with safety principle
- Add pre-commit verification step to confirm staged files
- Add pre-commit hook handling for the only exception where git add is permitted
- Add comprehensive error handling guidance with ABORT strategy