uzairansarii

Maintenance Knowledge Agent — Development skill for Claude Code

Development community

Retrieval-grounded maintenance assistant for garage staff: answers only from documents, cites the source, and says "I dont know" when it isn't covered.

How to install Maintenance Knowledge Agent

This entry records only its repository, not the path inside it, so there is no exact command to give. Open uzairansarii/maintenance-knowledge-agent and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

What Maintenance Knowledge Agent does

Retrieval-grounded maintenance assistant for garage staff: answers only from documents, cites the source, and says "I dont know" when it isn't covered. Copilot Studio + SharePoint, with a Python (TF-IDF + Claude) comparison.

Alternatives in Development

Haft — Engineering decisions engine that know when they're stale 1.4k ★
Claude Code Everything You Need To Know — Ultimate all-in-one guide to mastering Claude Code 1.3k ★
AgentSkillOS — Build your agent from 200,000+ skills via skill RETRIEVAL & ORCHESTRATION 373 ★

README

Maintenance Knowledge Agent

A chat assistant for garage staff. A mechanic or driver asks a question such as "What is the brake inspection interval for a tractor unit?" or "What do I do if I find a hydraulic leak?". The agent searches the maintenance documents, answers in plain language, and names the document and section the answer came from. If the documents do not cover the question, it says so and does not guess.

This is retrieval-grounded question answering: answers come only from a controlled set of documents, with citations, and "I don't know" is a valid answer.

All 10 maintenance documents in this repo are original training documents written for this project. Figures in them are illustrative and are **not** official maintenance guidance. No confidential company material is used.

What is in this repo

Part	Tool	Status
Agent (main deliverable)	Microsoft Copilot Studio	Instructions, settings and build guide included. Build and test in your own tenant: see `copilot-studio/`
Knowledge base	SharePoint document library	10 documents in `docs/docx/` (source in `docs/md/`)
Logging action	Power Automate	Flow spec and list schema in `copilot-studio/power-automate-flow.md`
Comparison prototype	Python, scikit-learn, Anthropic API	Working code in `python-prototype/`
Test set	20 questions with expected answers	`test/test_questions.json` and `.csv`
Tracking	Excel	`tracking/test-tracker.xlsx` (scores, summary, prompt change log)
Guides	Markdown	`guides/user-guide.md`, `guides/demo-script.md`

How it works

flowchart LR
    U[Mechanic / driver] -->|question| A[Agent]
    A -->|search| K[(Maintenance documents)]
    K -->|relevant sections| A
    A -->|answer + document and section| U
    A -.->|not in documents| L[Power Automate: log question]
    L --> S[(SharePoint list: Unanswered Questions)]