Skip to content

fix: improve agent/command/skill accuracy and update deprecated patterns#139

Merged
shinpr merged 6 commits intomainfrom
fix/sync-ccw-improvements
Mar 19, 2026
Merged

fix: improve agent/command/skill accuracy and update deprecated patterns#139
shinpr merged 6 commits intomainfrom
fix/sync-ccw-improvements

Conversation

@shinpr
Copy link
Owner

@shinpr shinpr commented Mar 19, 2026

Summary

  • Fix incorrect tool name: Replace "Task tool" with "Agent tool" (the actual Claude Code tool name) across all command files to prevent first-attempt invocation failures
  • Update deprecated testing patterns: MSW v1 → v2 (http/HttpResponse), fireEventuserEvent.setup() per RTL best practices
  • Convert "Prohibited Actions" to "Output Self-Check": Positive verification checklists are more effective than negative constraint lists for guiding LLM behavior (6 agents)
  • Clarify orchestrator delegation protocol: Explicit Agent tool reference, permitted tools whitelist, sequential execution ordering, and as-is data bridging rules
  • Add affectedLayers field: requirement-analyzer output now includes ["backend", "frontend"] for fullstack workflow support
  • Clean up skill descriptions: Remove internal implementation details from skill-optimization and documentation-criteria
  • Add test design patterns section: Port good/bad test examples to EN, reorder good-first to avoid LLM pattern imprinting
  • Bump version: 1.17.0 → 1.17.1

All changes applied symmetrically to EN and JA.

Test plan

  • Verify no "Task tool" or "Taskツール" references remain in commands
  • Verify no "Prohibited Actions" or "禁止事項" sections remain in agents
  • Verify MSW v2 API (http, HttpResponse) in frontend-typescript-testing skill
  • Verify userEvent.setup() pattern replaces fireEvent in test examples
  • Verify EN/JA content parity

🤖 Generated with Claude Code

shinpr and others added 6 commits March 19, 2026 14:01
Port non-plugin-specific improvements from claude-code-workflows PRs:
- Replace "Task tool" with "Agent tool" (correct tool name) in all commands
- Update MSW v1→v2 (http/HttpResponse) and fireEvent→userEvent in frontend testing skill
- Convert "Prohibited Actions" to "Output Self-Check" checklists in 6 agents
- Add explicit "Agent tool" and permitted tools reference to orchestrator protocols
- Add affectedLayers field to requirement-analyzer output for fullstack support
- Clean up internal details from skill descriptions (skill-optimization, documentation-criteria)

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Japanese translation of EN changes from previous commit:
- Replace "Taskツール"/"Task tool" with "Agentツール" in all commands
- Update MSW v1→v2 (http/HttpResponse) and fireEvent→userEvent in frontend testing skill
- Convert "禁止事項" to "出力セルフチェック" checklists in 6 agents
- Add explicit "Agentツール" and permitted tools reference to orchestrator protocols
- Add affectedLayers field to requirement-analyzer output for fullstack support
- Clean up internal details from skill descriptions (skill-optimization, documentation-criteria)

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Port good/bad Counter examples from JA version to resolve
pre-existing content gap between EN and JA.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Put correct pattern before avoid pattern to prevent LLM pattern
imprinting. Rename "Anti-patterns"/"アンチパターン" to
"Test Design Patterns"/"テスト設計パターン".

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Code comments are naturally English; no need to translate.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
@shinpr shinpr self-assigned this Mar 19, 2026
@shinpr shinpr merged commit 01c2a17 into main Mar 19, 2026
1 check passed
@shinpr shinpr deleted the fix/sync-ccw-improvements branch March 19, 2026 05:26
@shinpr shinpr mentioned this pull request Mar 19, 2026
2 tasks
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant