Agent Skills › repowise-dev/repowise › codebase-exploration

codebase-exploration

GitHub

指导 Agent 利用 Repowise 智能层探索和理解代码库。通过 get_overview、get_answer 等工具获取架构、调用关系及风险，替代传统文件读取，高效回答代码逻辑与结构问题。

plugins/claude-code/skills/codebase-exploration/SKILL.md repowise-dev/repowise

Trigger Scenarios

了解代码库整体架构解释特定模块或功能的工作原理查找符号的调用者和影响范围分析代码变更的风险

Install

npx skills add repowise-dev/repowise --skill codebase-exploration -g -y

More Options

Non-standard path

npx skills add https://github.com/repowise-dev/repowise/tree/main/plugins/claude-code/skills/codebase-exploration -g -y

Use without installing

npx skills use repowise-dev/repowise@codebase-exploration

指定 Agent (Claude Code)

npx skills add repowise-dev/repowise --skill codebase-exploration -a claude-code -g -y

安装 repo 全部 skill

npx skills add repowise-dev/repowise --all -g -y

预览 repo 内 skill

npx skills add repowise-dev/repowise --list

SKILL.md

Frontmatter

{
    "name": "codebase-exploration",
    "description": "Use when exploring, understanding, or answering questions about a codebase that has Repowise indexed (a .repowise\/ directory in the project root). Activates for \"how does X work\", \"explain the architecture\", \"where is Y implemented\", \"what does this module do\", or any task that needs an understanding of structure before diving into source files.\n",
    "user-invocable": false
}

Codebase Exploration with Repowise

This project has a Repowise intelligence layer. For an exact identifier, string or file name, Grep, Glob and Read are faster. For how code works, who calls it, what a change breaks and why it is shaped this way, reach for the Repowise MCP tools: they return documentation, ownership, history, decisions, and graph structure that plain file reads don't, usually in one round-trip instead of many.

Which tool for which question

You want…	Call
An exact identifier, string or file name	Grep, Glob and Read.
First orientation in an unfamiliar repo	`get_overview()`: architecture summary, key modules, entry points, git health, knowledge map. Skip it once you have the map.
How X works or flows end to end	`get_answer(question="how does …")`, one question per call: synthesised answer with citations + one `confidence` trust signal. Collapses the search → read → reason loop.
Callers before a rename or signature change	`get_context(targets=["<symbol>"], include=["references"])`: definition, imports and calls; a `complete` list is the whole edit set.
What breaks if X changes	`get_risk(targets=["<file>"])`.
Why X is shaped this way, before changing a pattern	`get_why(query="why …", targets=["<file>"])`.
Before finishing a multi-file change	`get_change_risk(revspec="<base>..HEAD")`, then `get_health(targets=["<files you touched>"])`.
A fuzzy concept, or symbol ids and line bounds to feed `get_symbol`	`search_codebase(query="…")`: hybrid search. `mode="auto"` routes an identifier to indexed symbol hits (`symbol_id`/line bounds → pipe into `get_symbol`), a path to file pages (→ `get_context`), and prose to semantic wiki search (each hit reports `search_method`: `embedding` vs `bm25`). Force a branch with `mode=symbol\|path\|concept\|hybrid`; narrow symbols with `symbol_kind`.
A triage card for specific files/symbols	`get_context(targets=[…])`: title, summary, signatures, hotspot bit, top callers, decision titles, symbol_ids (a row without one is `path::name`). Batch many targets in one call.
The actual source of one symbol	`get_symbol("path/to/file.py::Name")`: exact bytes with line bounds. Cheaper than Read + offset math. Use a `symbol_id` from `get_context`.

Recommended flow

New area you don't know → get_overview() once.
A specific question → get_answer(question=…) first.
- High confidence → answer it, cite the paths.
- medium/low confidence → follow best_guesses[0].file or fallback_targets[0] into get_context, then get_symbol for bytes.
A named symbol or a path you only need to find → Grep, Glob or Read. When you need its callers, risk or history, pass it to get_context, get_risk or get_why instead.
More files around a concept → search_codebase, then get_context on the hits (batched), then get_symbol only for the bodies you actually need.

Read source directly for details the indexed context doesn't cover.

Trust signals — verify when

_meta.stale_warning is present (the index has diverged from HEAD), or
confidence is medium/low, or
a result's search_method is bm25.

Otherwise the response is current — act on it.

Error handling

"No repositories found. Run 'repowise init' first." → suggest /repowise:init. Add --no-editor-setup if this repo is a scratch clone, a fixture, or a worktree, so init writes nothing outside .repowise/ (no editor config, agent hooks or git hook).
MCP tools unavailable → prefer the matching CLI slash commands when the plugin is installed (/repowise:ask, /repowise:context, /repowise:symbol, /repowise:search) rather than grepping blind.
get_answer/search_codebase come back empty → the repo may have a template-rendered wiki. Fall back to get_context with explicit paths, and note that model-written pages (repowise generate, or /repowise:init with an LLM provider) unlock richer docs + semantic search.
Tools fail to connect at all → the repowise binary may not be installed; suggest /repowise:init.

Version History

7ccd63f Current 2026-11-03 00:02
新增精确标识符查询路由至 Grep/Glob/Read；优化任务分流策略以匹配生成的文档规范。
09471b8 2026-10-19 07:00
优化 get_answer 的信任信号机制，统一使用 confidence 替代 retrieval_quality，并增强答案与问题的词覆盖率匹配逻辑；更新缓存 schema。
8f69324 2026-09-22 12:16
修复MCP无索引提示中的主机特定工具名称（如Read/Grep），改为通用能力描述以提升跨客户端兼容性。
370793f 2026-08-19 23:15
新增 CLI 适配器作为斜杠命令，并添加 --no-editor-setup 选项以跳过全局 MCP/钩子注册。
b5f4753 2026-07-25 05:12