ワンクリックで
prompt-patterns
Apply Codex-native prompt patterns for agents, skills, rules, and project instructions
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
メニュー
Apply Codex-native prompt patterns for agents, skills, rules, and project instructions
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
Canonical validation checklist for generated team structures, executed by a fresh-context verifier
Entry point that makes this session adopt the Team Architect workflow to run the full team design process
Start the A-Team Codex coordinator workflow for designing or restructuring teams
Rewrite generated prompts to be sharper, shorter, and more executable without changing meaning
Validate that a generated Codex team is structurally complete and internally consistent
Provide systematic prompt optimization methodology to improve instruction quality while preserving original characteristics
SOC 職業分類に基づく
| name | Prompt Patterns |
| description | Apply Codex-native prompt patterns for agents, skills, rules, and project instructions |
Use this skill when writing or reviewing prompts for Codex teams. It converts broad behavioral wishes into structured, testable prompt sections.
.codex/agents/generation/agent-writer.md.codex/agents/generation/skill-writer.md.codex/agents/generation/rule-writer.md.codex/agents/optimization/prompt-optimizer.md## Boundaries for exclusions instead of relying on "do not" sentences.## Input and Output for handoff contracts.## Completion Contract for status, artifacts, changed files, worklog paths, and blockers.## Preflight and ## Verification for concise decision notes and quality checks.## Uncertainty Protocol for INSUFFICIENT_DATA, NEEDS_CONTEXT, and BLOCKED.AGENTS.md for project-wide runtime guidance..codex/config.toml and agents/**/*.toml for subagent registration and runtime behavior..agents/skills/{skill-name}/SKILL.md for runtime-discoverable skills..codex/rules/{rule-name}.md for enforceable constraints..claude/settings.json, Task tool, Agent tool, context: fork, and Claude model names.When examples are included, cover:
Make the reviewer careful and evidence-based.
Add ## Evidence Requirements, ## Verification, and ## Completion Contract sections. Require each finding to cite a file path or worklog entry, and require INSUFFICIENT_DATA when evidence is missing.