メニュー
Navigate and understand codebase structure using the code-review graph. Use for repo orientation, AGENTS.md authoring, architecture mapping, or finding relevant code.
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。
SOC 職業分類に基づく
Use when asked to run a full repo hygiene pass, full cleanup, or to maintain a repo in one go — and whenever the user invokes /safe-code or any wrapper of it (/skill:safe-code, /skills safe-code, $safe-code, @safe-code, or bare safe-code), including --continue to resume saved work and --save to finalize docs and commit. Also use for first-time project setup, restoring project context or session memory, dead-code audits, or agent-config trust checks.
Scan an entire codebase to detect and safely remove dead code such as unused functions, orphaned modules, unreferenced exports, stale configs, dead routes, and leftover workflow artifacts. Use when asked to clean up dead code, remove unused code, prune stale files, find orphaned modules, audit codebase bloat, or delete code left behind after a workflow or architecture change.
Refactor code safely in small verified slices while keeping repo continuity docs in sync. Uses code-review graph tools for rename previews, impact radius, affected flows, and post-change review when available. Use when an agent is asked to refactor, restructure, clean up, remove or replace code, modernize modules, or do follow-up hygiene in a repo.
Senior engineer discipline layer for any coding task. Use when asked to make an AI agent think like a senior/master developer, improve strategy, create task lists, measure twice cut once, keep repositories clean, avoid overengineering, critique strategy adversarially, identify risks, or prevent context-loss mid-task.
Build or update the code review knowledge graph. Use before safe-code audits, refactors, reviews, debugging, or when the graph may be stale.
Systematically debug issues using graph-powered code navigation, recent-change detection, affected flows, and targeted verification.
| name | explore-codebase |
| description | Navigate and understand codebase structure using the code-review graph. Use for repo orientation, AGENTS.md authoring, architecture mapping, or finding relevant code. |
Use the code-review graph to understand the repository before editing.
get_minimal_context_tool(task="<your task>").detail_level="minimal" unless more detail is needed.rg, manifests, README files, and config inspection.get_minimal_context_tool.get_architecture_overview_tool() for major boundaries.list_communities_tool(detail_level="minimal") to identify modules.list_flows_tool(detail_level="minimal") for execution paths.get_hub_nodes_tool() and get_bridge_nodes_tool() for risky chokepoints.semantic_search_nodes_tool() or query_graph_tool() to narrow to specific symbols.When helping safe-code write or reconcile AGENTS.md, follow the canonical authoring rules in the safe-code skill's references/agents-md-authoring.md (decision test, investigation order, what to extract/exclude). This skill's job is to supply graph-backed facts that feed those rules, not to define a separate authoring process.
Prefer graph-backed facts:
Do not copy raw graph dumps into docs. Convert them into compact handoff facts.