mcpField
K
knowledge-base-query

knowledge-base-query

github.com/mmontes11Productivity174 installsUpdated 2w ago
Repository

About knowledge-base-query

A read-only skill for answering questions from a repository-based knowledge base covering projects, books, and foundational topics. It uses indexes to locate only the relevant documents and grounds responses in those sources.

Key features

  1. 01Find entries through repository indexes
  2. 02Read only the relevant documents
  3. 03Answer with knowledge-base citations
  4. 04Surface upstream links for thin or stale entries
  5. 05Keep knowledge-base content unchanged

Use cases

  1. 01Checking a project's documented features
  2. 02Finding book or chapter information
  3. 03Answering questions about foundational topics

Install

Copy the command and run it in a terminal.

$npx skills add https://github.com/mmontes11/skills --skill knowledge-base-query

After installing, type/knowledge-base-queryin Claude Code to confirm it loaded.

Details

From the project README

Query the knowledge base

Read information from the knowledge base — projects, books, and foundational topics — and answer grounded in it. The root README.md is the index: it maps names to folders, so start there and go straight to the target. Do not read the whole repository; use the indexes to open only the file(s) the question needs.

Locate the knowledge base

Work inside the knowledge-base repository. The canonical local checkout is ~/code/knowledge-base; if it is absent, git clone [email protected]:mmontes11/knowledge-base.git ~/code/knowledge-base. All paths below are relative to the repository root.

Step 0 — Find it in the index

The root README.md holds the indexes:

  • Tracked projects table — maps each project to its projects/<category>/<project>/ folder and its upstream URL.
  • Tracked books table — maps each book to its books/<book>/ folder.
  • Tracked foundational knowledge table — maps each foundational topic to its foundation/<topic>/ folder.

If the thing you were asked about is not in the index, say so — the knowledge base does not cover it — and stop. If a Tracked projects row exists and further research would help, surface its upstream URL.

Step 1 — Read the right file

Go straight to the file that answers the question:

QuestionRead
What is the project / what problem does it solveprojects/<c>/<p>/README.md
What is its API / resource / CLI surfaceprojects/<c>/<p>/api-reference.md
What is the latest release / what changedprojects/<c>/<p>/releases.md
What features does it haveprojects/<c>/<p>/features.md
How is it architectedprojects/<c>/<p>/architecture.md (only if present)
What a book teaches / a chapter's contentbooks/<book>/README.md, then NN-<chapter>.md
How a concept actually worksfoundation/<topic>/README.md

Step 2 — Answer

  • Answer from the knowledge base, citing the file you read and, where useful, its last_updated frontmatter so the reader knows how current it is.
  • For projects the KB links rather than copies upstream. If the entry is thin or stale, say so and surface the upstream link from the Tracked projects table so the reader can verify.
  • If the question spans several files (e.g. "give me the picture of "), read README.md first, then pull in only the specific files the question needs.

Scope

This skill is read-only — it never edits, adds, or removes knowledge-base content. Adding a project or a book, and refreshing a project, belong to the knowledge-base-add-project and knowledge-base-add-book skills.

Comments

Markdown supported