CodeQuest — drop-in repo intelligence for GitHub Copilot

Copy this folder into any repository. It turns GitHub Copilot (agent mode) into CodeQuest Core AI: a repo-comprehension engine that maps codebases, computes blast radius, narrates execution flows, simulates breaking changes, and gates architecture drift in CI — grounded in deterministic static analysis, not guesses.

60-second quickstart

# 1. Drop this folder into your repo (or clone CodeQuest and copy ./codequest)
cp -r codequest /path/to/your-repo/

# 2. Scan the repo — zero dependencies, runs offline (Node 18+)
cd /path/to/your-repo && ./codequest/bin/codequest scan

# 3. Auto-configure your IDEs — detects VS Code, Cursor, Windsurf, Zed, …
./codequest/bin/codequest setup

# 4. Optional: set your audience for explanations
./codequest/bin/codequest init --audience professional   # or beginner | expert
./codequest/bin/codequest init --audience expert --business  # non-technical lens

# 5. Open your IDE's AI agent and run:
#    /codequest-onboard

That's it. Copilot reads CODEQUEST.md + .codequest/graph.json and becomes the CodeQuest AI.

What's inside

| Path | What it is | |---|---| | CODEQUEST.md | The master system prompt — the brain. | | .github/muse-instructions.md | Auto-loaded repo instructions for Copilot. | | .github/instructions/ | Path-scoped analysis + formatting rules. | | .github/prompts/ | One-shot slash prompts: /codequest-onboard, /codequest-map, /codequest-blast-radius, /codequest-story, /codequest-whatif, /codequest-review. | | skills/ | Agent Skills (repo-cartographer, blast-radius, story-mode, what-if) for Copilot agent mode. | | bin/codequest | Deterministic static analyzer (Node, zero deps). Builds the module graph Copilot reasons over. | | mcp-server.js + mcp.example.json | The analyzer as MCP tools — Copilot calls them natively; add your company's APIs alongside. See docs/MCP.md. | | templates/ | Report templates: blast-radius, story-mode, module-card, architecture-overview. | | AGENTS.md | Agent operating notes (also read by Copilot). |

The loop

./codequest/bin/codequest scan        # deterministic: modules, imports, graph, risk
        │
        ▼
Copilot agent mode (/codequest-*)     # intelligence: summaries, narratives, judgment
        │
        ▼
.codequest/ + Markdown reports        # saved in-repo: works offline, diffable, reviewable

Commands

./codequest/bin/codequest scan [dir]            # build .codequest/graph.json + repo-map.md
./codequest/bin/codequest map [dir]             # print the repo map
./codequest/bin/codequest blast-radius <file>   # impact matrix for a module
./codequest/bin/codequest story <entry-file>    # static execution trace from an entry point
./codequest/bin/codequest modules [dir]         # list modules + exports
./codequest/bin/codequest diff <old> <new>      # architecture drift gate (CI: fails on new CRITICAL/cycles)

Add --json to any command for machine-readable output.

Requirements

• Node.js 18+ (analyzer only; no npm install needed — zero dependencies)
• VS Code + GitHub Copilot (agent mode) for the intelligence layer

Privacy

The analyzer never touches the network. Everything it produces lives in .codequest/ inside your repo. Copilot interactions stay inside your existing Copilot session — no new services, no code exfiltration beyond what you already trust Copilot with.