AI 작업 지침을 Notion으로 이전하고 DevLog_Harness를 제거한다

.codex/agents/architecture_watcher.toml

@@ -4,8 +4,6 @@ model = "gpt-5.3-codex-spark"
 model_reasoning_effort = "xhigh"
 sandbox_mode = "read-only"
 developer_instructions = """
-Read AGENTS.md, README.md, .agents/rules/general.md, and .agents/rules/architecture.md before analysis.
-Inspect only the architecture scope assigned by the main agent. Do not edit files or change GitHub state.
-Do not run, launch, install, boot, or open the app or Simulator.
-Return a concise verdict, changed boundary, evidence, and any user decision required.
+Read AGENTS.md and complete its Notion policy bootstrap, including the active Architecture policy, before analysis.
+Read README.md and follow the active iOS custom agent contract for this agent and the assigned task.
 """

.codex/agents/architecture_watcher_luna.toml

@@ -4,8 +4,6 @@ model = "gpt-5.6-luna"
 model_reasoning_effort = "xhigh"
 sandbox_mode = "read-only"
 developer_instructions = """
-Read AGENTS.md, README.md, .agents/rules/general.md, and .agents/rules/architecture.md before analysis.
-Inspect only the architecture scope assigned by the main agent. Do not edit files or change GitHub state.
-Do not run, launch, install, boot, or open the app or Simulator.
-Return a concise verdict, changed boundary, evidence, and any user decision required.
+Read AGENTS.md and complete its Notion policy bootstrap, including the active Architecture policy, before analysis.
+Read README.md and follow the active iOS custom agent contract for this agent and the assigned task.
 """

.codex/agents/code_reviewer.toml

@@ -0,0 +1,10 @@
+name = "code_reviewer"
+description = "Read-only DevLog reviewer used only when the user explicitly requests review in the current turn."
+model = "gpt-6-astra"
+model_reasoning_effort = "medium"
+sandbox_mode = "read-only"
+developer_instructions = """
+Read AGENTS.md and complete its Notion policy bootstrap, including every task-matching active policy, before review.
+Proceed only when the parent confirms that the user explicitly requested review in the current turn.
+Follow the active iOS code_reviewer contract for this agent and the assigned diff.
+"""

.codex/agents/documentation_writer.toml

@@ -4,8 +4,6 @@ model = "gpt-5.3-codex-spark"
 model_reasoning_effort = "xhigh"
 sandbox_mode = "workspace-write"
 developer_instructions = """
-Read AGENTS.md and all task-matching rules under .agents/rules before drafting.
-Match repository templates, the actual diff, live issue or PR state supplied by the main agent, and the requested Korean wording.
-Edit only documentation files explicitly assigned by the main agent. Do not edit app code or change GitHub state.
-Return the draft or changed paths, sources used, and unresolved decisions concisely.
+Read AGENTS.md and complete its Notion policy bootstrap, including every task-matching active policy, before drafting.
+Follow the active iOS custom agent contract for this agent and the assigned task.
 """

.codex/agents/documentation_writer_luna.toml

@@ -4,8 +4,6 @@ model = "gpt-5.6-luna"
 model_reasoning_effort = "xhigh"
 sandbox_mode = "workspace-write"
 developer_instructions = """
-Read AGENTS.md and all task-matching rules under .agents/rules before drafting.
-Match repository templates, the actual diff, live issue or PR state supplied by the main agent, and the requested Korean wording.
-Edit only documentation files explicitly assigned by the main agent. Do not edit app code or change GitHub state.
-Return the draft or changed paths, sources used, and unresolved decisions concisely.
+Read AGENTS.md and complete its Notion policy bootstrap, including every task-matching active policy, before drafting.
+Follow the active iOS custom agent contract for this agent and the assigned task.
 """

.codex/agents/github_ci_analyst.toml

@@ -4,8 +4,6 @@ model = "gpt-5.3-codex-spark"
 model_reasoning_effort = "xhigh"
 sandbox_mode = "read-only"
 developer_instructions = """
-Read AGENTS.md and .agents/rules/project-workflows.md before analysis.
-Use live GitHub state as the source of truth and inspect review threads when resolution state matters.
-Do not edit files, reply, resolve threads, submit reviews, push, or change GitHub state.
-Return concise current state, actionable items, evidence links, and unresolved decisions.
+Read AGENTS.md and complete its Notion policy bootstrap, including the active Project Workflows policy, before analysis.
+Follow the active iOS custom agent contract for this agent and the assigned task.
 """

.codex/agents/github_ci_analyst_luna.toml

@@ -4,8 +4,6 @@ model = "gpt-5.6-luna"
 model_reasoning_effort = "xhigh"
 sandbox_mode = "read-only"
 developer_instructions = """
-Read AGENTS.md and .agents/rules/project-workflows.md before analysis.
-Use live GitHub state as the source of truth and inspect review threads when resolution state matters.
-Do not edit files, reply, resolve threads, submit reviews, push, or change GitHub state.
-Return concise current state, actionable items, evidence links, and unresolved decisions.
+Read AGENTS.md and complete its Notion policy bootstrap, including the active Project Workflows policy, before analysis.
+Follow the active iOS custom agent contract for this agent and the assigned task.
 """

.codex/agents/verification_runner.toml

@@ -4,12 +4,6 @@ model = "gpt-5.3-codex-spark"
 model_reasoning_effort = "xhigh"
 sandbox_mode = "workspace-write"
 developer_instructions = """
-Read AGENTS.md and all task-matching rules under .agents/rules before verification.
-Run only the checks assigned by the main agent. Record exact commands, exit status, evidence, failures, and skipped checks.
-Compare the assigned checks with every required check for the changed file types in the repository rules.
-Return Status: Pass only when every required check ran and passed, Fail when a required check ran and failed, or Not Run when any required check was skipped or unavailable.
-Never report a skipped or unavailable required check as Pass.
-Do not edit source or documentation files except through an explicitly assigned formatting command.
-Do not run, launch, install, boot, or open the app or Simulator.
-Return a concise verification result.
+Read AGENTS.md and complete its Notion policy bootstrap, including every verification-related active policy, before verification.
+Follow the active iOS custom agent contract for this agent and the assigned task.
 """

.codex/agents/verification_runner_luna.toml

@@ -4,12 +4,6 @@ model = "gpt-5.6-luna"
 model_reasoning_effort = "xhigh"
 sandbox_mode = "workspace-write"
 developer_instructions = """
-Read AGENTS.md and all task-matching rules under .agents/rules before verification.
-Run only the checks assigned by the main agent. Record exact commands, exit status, evidence, failures, and skipped checks.
-Compare the assigned checks with every required check for the changed file types in the repository rules.
-Return Status: Pass only when every required check ran and passed, Fail when a required check ran and failed, or Not Run when any required check was skipped or unavailable.
-Never report a skipped or unavailable required check as Pass.
-Do not edit source or documentation files except through an explicitly assigned formatting command.
-Do not run, launch, install, boot, or open the app or Simulator.
-Return a concise verification result.
+Read AGENTS.md and complete its Notion policy bootstrap, including every verification-related active policy, before verification.
+Follow the active iOS custom agent contract for this agent and the assigned task.
 """