Skip to content

CHANGELOG.md

Latest commit

 

History

History
108 lines (78 loc) · 5.65 KB

CHANGELOG.md

File metadata and controls

108 lines (78 loc) · 5.65 KB

Changelog

Alle relevanten Änderungen an diesem Projekt werden hier dokumentiert.

Das Format basiert auf Keep a Changelog und dieses Projekt folgt Semantic Versioning.

[0.3.6] - 2026-04-19

Changed

  • auto-dev: Major upgrade to review pipeline, adopting github-issue-fixer patterns:
    • 4 parallel reviewers always (Logic, Security, Quality, Architecture) instead of 1-3 selected by task type
    • Dynamic reviewer participation — only reviewers with findings re-enter next round
    • 5-iteration budget (shared between review-fix loop and acceptance) instead of 3
    • New Phase 5: Test hardening with regression, edge-case, and error-path tests
    • New Phase 6: Acceptance review — all 4 reviewers verify code + hardening tests together
    • Report mode when iterations exhausted (no commit, findings documented) instead of simple keep/revert
    • Optional UI mockup generation in Phase 2 — served via localhost for user review
    • 3 new agents: architecture-reviewer, test-writer, mockup-designer
    • New finding-format.md with consolidation rules for 4 reviewers
    • Updated report template with hardening, acceptance, and report mode sections

[0.3.5] - 2026-04-19

Changed

  • github-issue-fixer: Complete redesign as 8-wave architecture:
    • Wave 1: Dual-agent issue validation (Analyzer + Validator with independent second opinion)
    • Wave 2: Enhanced planning with TDD test strategy
    • Wave 3-4: TDD flow (reproduction test first, then fix)
    • Wave 5: Iterative review-fix loop with 4 parallel reviewers (Logic, Security, Quality, Architecture) — max 5 iterations
    • Wave 6: Test hardening with regression and edge-case tests
    • Wave 7: Final acceptance review by all 4 reviewers
    • Wave 8: Commit with GitHub issue lifecycle management (comment + auto-close)
    • 10 agent definitions (3 modified, 7 new), 2 new reference files

[0.3.1] - 2026-04-12

Changed

  • auto-dev: Questions now include recommendations with reasoning. Questions are presented one at a time, waiting for user response before proceeding.

[0.3.0] - 2026-04-11

Added

  • auto-dev: New skill — universal autonomous task executor with adaptive questions, parallel implementation agents, and iterative review-fix loop (8 agent types, 6 phases).

[0.2.0] - 2026-04-09

Added

  • codebase-onboarding: New skill — analyzes codebases and generates ARCHITECTURE.md / GETTING-STARTED.md (3 agents).
  • test-engineer: New skill — finds coverage gaps and writes missing tests in 2 waves (5 agents).
  • pr-reviewer: New skill — multi-perspective PR review with 3 parallel agents (Logic, Security, Quality).
  • perf-analyzer: New skill — performance bottleneck analysis with smart agent selection (4 agents: Bundle, Query, Runtime, Infra).

[1.2.0] - 2026-04-09

Changed

  • All skills, agent prompts, references, and scripts translated from German to English.
  • SKILL.md description fields now include both English (primary) and German (secondary) trigger phrases.
  • Shell script comments and output strings translated to English.

[1.1.0] - 2026-04-09

Hinzugefügt

  • Shared references/agent-invocation.md — Standard für Agent-Start, Rückgabeformat und Fehlerbehandlung.
  • Shared references/finding-format.md — Vereinheitlichtes Finding-Format für alle Analyse-Agenten.
  • github-issue-fixer/agents/coder.md — Fehlender Agent für Phase 3 (Fix-Implementierung).
  • Datei-Partitionierungs-Algorithmus für codebase-doctor Wave 2 (Auto-Fix).
  • Loop-Limit (max 2 Iterationen) für codebase-doctor Wave 3 (Review).
  • Fixbar-Bewertung in alle 7 codebase-doctor Agenten integriert.

Geändert

  • refactor-orchestrator komplett in modulare Struktur aufgeteilt (4 Agent-Dateien + Report-Template).
  • /tmp/-basierte Inter-Agent-Kommunikation durch Markdown-Antworten ersetzt.
  • claude -p Referenzen durch Agent-Tool-Invokation ersetzt.
  • Severity-Emojis (🔴🟠🟡🟢) in codebase-doctor vereinheitlicht mit audit-project.
  • Severity-Bereiche in quality.md und api-consistency.md erweitert (bis high).
  • Terminologie vereinheitlicht: "Risiko" → "Auswirkung" in allen Agenten.
  • Rate-Limit in audit-project SKILL.md auf 2s angepasst (konsistent mit Script).

Behoben

  • disable-model-invocation: true in github-issue-fixer entfernt (blockierte Skill-Aufruf).
  • eval "$@" Sicherheitslücke in detect-and-run-tests.sh durch direkte Ausführung ersetzt.
  • Subshell-Variable-Bug in create-audit-issues.sh (Process Substitution statt Pipe).
  • Fehlerhafte JSON-Ausgabe in create-audit-issues.sh korrigiert.
  • Fehlerhafte Issue-Count-Berechnung in project-info.sh korrigiert.
  • Fehlerhafte Language-Array-Generierung in project-info.sh korrigiert.
  • Fehlende Exclude-Patterns in github-issue-fixer Analyzer-Agent hinzugefügt.

[1.0.0] - 2026-04-09

Hinzugefügt

  • audit-project: Umfassendes Projekt-Audit mit 5 parallelen Subagenten (Security, Bugs, Code-Hygiene, Struktur, GitHub Issues) und automatischer GitHub Issue-Erstellung.
  • codebase-doctor: 3-Wellen Codebase-Analyse (7 Analyse-Agenten → Auto-Fix → Review & Verify) mit automatischer Fehlerbehebung.
  • github-issue-fixer: Systematischer GitHub Issue Fix-Workflow (Analyse → Plan → Fix → Verify → Commit) mit optionaler Browser-Verifikation via MCP DevTools.
  • refactor-orchestrator: Multi-Agent Refactoring mit Teamleader-Koordination und parallelen Scout/Architekt/Worker/Test-Agenten.
  • Plugin-Manifest (.claude-plugin/plugin.json) für die Installation als Claude Code Plugin unter dem Namen codewright.
  • README.md mit Installationsanleitung und Skill-Dokumentation.
  • CLAUDE.md für Claude Code Kontext.