القائمة
Core guidance for RepoQL - when to use it vs raw file reads, and the explore workflow.
التثبيت باستخدام Codex أو Claude انسخ هذا Prompt والصقه في Codex أو Claude أو مساعد آخر ليراجع صفحة Skill ويثبّتها لك.
استنادا إلى تصنيف SOC المهني
Diagnose and fix RepoQL issues. Use when a RepoQL tool call fails or errors, results seem wrong or incomplete, indexing or queries are slow, a directory isn't being indexed or isn't recognized as a workspace, an imported repo won't appear, semantic search or explain returns nothing, or the host is unresponsive.
How to write clear, verifiable, discoverable documents for LLMs and humans. Always use when writing north-star, gestalt, reference, research, design, plan, flow, findings, concepts, or process documents.
Effective Mermaid diagrams. Use when creating markdown documents with human audiences.
Search patterns for RepoQL - semantic search, scoping, and question reranking.
Build a Claude Code status line powered by RepoQL — blend the session transcript with the indexed graph (cost, cache, tools, index state, dashboard link) and avoid the footguns that make a status line break or lie. Use when writing or extending a status-line script.
Effective use of RepoQL — the workflow, the techniques, and the wild magic. Use when you want to get more out of RepoQL or need to understand what's possible.
| name | repoql |
| description | Core guidance for RepoQL - when to use it vs raw file reads, and the explore workflow. |
RepoQL is a local knowledge graph for repositories. It indexes code structure, enabling exploration without reading every file.
Use RepoQL when:
Use regular file reads when:
Use keywords as the vocabulary probe, and add question when you have a specific intent. question reranks results toward the answer; it does not replace keywords.
Typical progression:
keywordskeywordsrepoql_readrepoql_explain| Tool | Use Case |
|---|---|
repoql_explore | Discovery and understanding (start here) |
repoql_keywords | Reshape rough terms into the repository's real vocabulary |
repoql_query | SQL aggregation and filtering over the graph |
repoql_read | Fetch content with token budget (URIs, fragments, modifiers) |
repoql_explain | Synthesized answers with citations |
repoql_execute | Sandboxed JavaScript over the graph — diagrams, conversions, artifacts |
repoql_import | Add or remove external repositories |
repoql_capture_concept | Write a durable invariant into the repo's concept memory |
repoql_command | Management commands — config, diagnostics, account, host, imports |
repoql_watch | Run a process under the host OTEL collector and query its telemetry |
repoql_status | Check host/socket/plugin health |
The plugin exposes the same surface as the RepoQL MCP server; each tool's
description carries the full MCP guidance. repoql_explore and
repoql_keywords are the discovery entry points; repoql_query and
repoql_execute are the power tools; repoql_command is the management remote.
Budget = tokens you want to spend on the response.
| Task | Typical Budget |
|---|---|
| Inventory | 800-2000 |
| Locate | 1000-2000 |
| Inspect | 2000-5000 |
| Explain | 1000-3000 |
Higher budget = richer detail. Start small, increase if needed.