this file to guide all Copilot/AI code generation in this project.**
---
## Docker MCPServer Naming Advice
- When adding a Dockerized MCPserver, **always use named containers** by including the `--name
016mo agoReviewMountCreate a free account to mount
Engram — AI Agent Instructions
Engram is a local MCPserver providing persistent memory for AI coding agents. It exposes 4 dispatcher tools (`engram_session`, `engram_memory`, `engram_admin`, `engram_find
runtime commands used in this repo:
```bash
# deterministic offline eval
.venv/bin/python evals/run_offline_eval.py
# run MCPserver locally (stdio)
.venv/bin/anthropic-news-mcp
# or
.venv/bin/python -m anthropic_news_mcp
```
## High-level architecture
The server
registered in `conftest.py` + `pyproject.toml`'s `[tool.setuptools.package-dir]`.
- Tests are in `16_tests/`.
- The MCPserver lives at `02_core/interfaces/mcp/` (registered as `core.interfaces.mcp`).
## Key Conventions
- Use clean import names (numeric prefixes
/elado` prompt file (`.github/prompts/elado.prompt.md`) for
the full workflow, and the `magyar-elado` MCPserver (`.vscode/mcp.json`) for
live price data.
Core rules:
- Never guess prices — research comparables (MCP `search_market_prices
skills/yjs-best-practices/SKILL.md` for Yjs data modeling, updates,
providers, performance, and debugging questions.
- Use the Liveblocks MCPserver from `.mcp.json` only with a development project
secret key. Do not use a production
knows the reviewer number (`EXPO_PUBLIC_REVIEWER_PHONE` in `src/store/authStore.ts`); the code is checked server-side by the phone-session function (`REVIEWER_PHONE` / `REVIEWER_OTP`) and must never
Read: `MilestoneSmoke/post-milestone-smoke-test.ios.prompt.md`
- **post-milestone-smoke-test-web** — Drive a web app through a browser MCPserver after a milestone tag is cut — exercise headline flows, write a pass/fail/blocked report.
Read
sequence of reads, searches, and edits, check whether an available skill, slash command, or MCPserver already performs this task in one purpose-built step — it's usually both cheaper
duplicate root-level instruction files.
- Do not invoke the ADO or Bluebird MCPservers for work in this repository,
even when they are available; they do not provide useful Kusto
Control Repository Instructions
This repository is `computer-control-mcp`, a cross-platform C++20 MCPserver for desktop and mobile-simulator automation. The MCPserver is the primary contract
Game Discovery MCPServer - AI Agent Instructions
## Architecture Overview
This is a **Model Context Protocol (MCP) server** built as a **single C# script** (`game-mcp.cs`) that discovers installed games across multiple
skill-awareness — resolving dependency trees, wiring skills into agent directories, and configuring MCPservers.
**Design principle: stay as close to npm as possible.** skillpm should feel like npm. Same `package.json
loop for anything irreversible; headless/scheduled runs are read/review-only by default; never assume an MCPserver exists; document the why. Full rules: .claude/CLAUDE.md §8.
- Specs precede code. Write small, safe, reversible
server (tool catalogue, decision tree, token economy, error codes):
> [`docs/AGENT_CONTRACT.md`](../docs/AGENT_CONTRACT.md).
## Overview
MCPserver bridging pkinative 1.0.0 (pinned `^1.0.0`) to AI clients over stdio or Streamable
HTTP, on the MCP