AGENTS.md is an exact copy of this file; make check-agents-sync keeps the two in sync. Run a single unit test: Run a single integration test case: Update golden files after intentional changes. Unit and integration golden files regenerate separately: UAT tests validate rendered catalog output against golden files using a Kind cluster with KWOK-simulated GPU nodes. Requires: Docker, Kind, Tilt, Go 1.24+. For development iteration (faster feedback loop): Run a single UAT test: UAT golden files are in test/uat/testdata/<csp>/<gpu>/nccl/expected_pods.yaml. When…
8433d agoReviewMountCreate a free account to mount
Guidance for Claude Code when working with this repository. Instructions here override default behavior. Three tenets govern every change in this repo — to code, MCP tools, or prompts. Read docs/PHILOSOPHY.md (mirrored as mcp-steroid://skill/design-philosophy for runtime fetch via steroid_fetch_resource) before proposing any of: Short version: the MCP tool surface (8 today) stays narrow on purpose; the IntelliJ capability surface stays full, exposed via steroid_execute_code plus prompt resources. The strategy page's "Give AI the whole IDE, not just the files" is…
823mo agoReviewMountCreate a free account to mount
This folder holds long-form research, design plans, and the DPAIA arena working log. Read this in addition to the root CLAUDE.md when changing files under docs/ or referencing autoresearch results. The repo's four design tenets — small MCP tool surface; power lives in prompts and direct IntelliJ API usage; devrig is stateless; McpScriptContext is last-resort — are codified in PHILOSOPHY.md (and mirrored at runtime as mcp-steroid://skill/design-philosophy). Autoresearch and prompt- optimization work in this folder is the primary lever for those…
823mo agoReviewMountCreate a free account to mount
Guidance for working on the IntelliJ Platform plugin and standalone MCP server. Read this in addition to the root CLAUDE.md whenever you change files under ij-plugin/. IntelliJ Platform plugin with standalone MCP server (Kotlin MCP SDK + Ktor): Core flow: SteroidsMcpServer.kt → ExecuteCodeToolHandler.kt → ExecutionManager.kt → CodeEvalManager.kt (compile) → ScriptExecutor.kt (run with timeout) Key decisions: Standalone MCP server (no built-in MCP dependency). Synchronous request-response. Coroutines over blocking. Two-phase execution (compile then run). Fast failure on errors. Disposable lifecycle for cleanup.
823mo agoReviewMountCreate a free account to mount
Guidance for the :installer-gen module. Instructions here override the root guide for this folder. Read the root CLAUDE.md too (project-wide rules). :installer-gen is build-tooling (no IntelliJ deps). It is the lower of the two distribution generators: :website-gen depends on it. Two responsibilities: 1. JDK detection — resolveAllJdks(cache) (in JdkModel.kt) produces a JdkModel of the JDK 25 builds the installer ships, each field computed from the live vendor sources (nothing hand-pinned). The vendor logic is split per file: CorrettoJdk.kt (Amazon Corretto)…
823mo agoReviewMountCreate a free account to mount
Experimental / long-running / less stable Docker-based tests. Read this in addition to the root CLAUDE.md and test-integration/AGENTS.md (the shared infrastructure lives there) when changing files under test-experiments/. The three repo-wide design tenets (docs/PHILOSOPHY.md; runtime mirror mcp-steroid://skill/design-philosophy) apply here too — arena prompts and debugger demos are exactly the surface that "agents deliver more" measurements run against, so prompt-quality wins land here. New steroid_* tools and new McpScriptContext methods do not. The following tests are in :test-integration even though they…
823mo agoReviewMountCreate a free account to mount
Cross-OS behaviour tests (Windows and Linux) for the script-only, minimal-dependency way an agent plugin launches things — its stdio MCP server and its hooks — validating how Claude Code resolves each candidate script form per OS. Uses mock probe scripts (no real devrig). Live counterpart to jonnyzzz/mcp-steroid#253. The suite is structurally tied to a real Windows/Linux agent (it downloads + runs the OS Claude build). Per the root CLAUDE.md rule, it is gated at the task level: build.gradle.kts → tasks.test…
823mo agoReviewMountCreate a free account to mount
This directory contains the Hugo source for the MCP Steroid website at https://mcp-steroid.jonnyzzz.com Scope note: the public website targets end users — install instructions, demo videos, the strategy page. Agent-facing design tenets live in docs/PHILOSOPHY.md (canonical, mirrored at runtime as mcp-steroid://skill/design-philosophy); contributor guidance for the plugin / prompts / tests lives in the per-folder CLAUDE.md and AGENTS.md files. Site copy and agent docs evolve on different cadences and to different audiences; don't conflate them. The website is built from this…
823mo agoReviewMountCreate a free account to mount
Guidance for the :website-gen module. Instructions here override the root guide for this folder. Read the root CLAUDE.md too (project-wide rules). :website-gen is build-tooling, not plugin runtime — no IntelliJ dependencies. One job: - Website artifacts (WebsiteArtifacts.kt, task generateWebsite) — resolves the published GitHub release and writes version.json + updatePlugins.xml (the IntelliJ custom-plugin-repository XML) into website/build/generated-static, and copies EULA + LICENSE there too. The task knows all paths (VERSION + release notes from the project layout); CI/Makefile invoke it with…
823mo agoReviewMountCreate a free account to mount
The canonical file lives at .claude/CLAUDE.md. The root-level AGENTS.md is a symlink to this file so agents that expect the cross-agent AGENTS.md convention find the same content. Skyhook (being renamed to NodeWright) is a Kubernetes-aware package manager for safely modifying host infrastructure at scale. It coordinates the node lifecycle (cordon → drain → apply package → interrupt/reboot → uncordon) as controlled rollouts gated by interruption budgets and deployment policies. Rename status: the project is transitioning from Skyhook → NodeWright. Components…
See AGENTS.md and .jules/sentinel.md for full repository architecture, mandatory review guardrails, and verification steps. See AGENTS.md for OpenWiki agent instructions.
This workspace may contain up to 6 related repositories. Not all teams have all repos. Always be aware of which repository you're in when making changes. What is omnichannel-chat-widget? This is a React component library providing pre-built, customizable UI components for chat experiences. It includes message bubbles, input boxes, file attachments, typing indicators, and more. Components are theme-able and accessible (ARIA support). Key Features: - Pre-built React chat components - Theme customization support - Accessibility (ARIA attributes, keyboard navigation) -…
745mo agoReviewMountCreate a free account to mount
A catalog of Claude Code skills for Elastic documentation workflows. Skills are installed into ~/.claude/skills/ via install.sh and used as slash commands or auto-triggered behaviors in Claude Code. Every skills/**/SKILL.md must have YAML frontmatter: Agentic workflows use .md files compiled to .lock.yml via gh aw compile. The lock files are auto-generated — never edit them directly. The .gitattributes file marks them as linguist-generated=true merge=ours. Use /create-skill within this repo, or see CONTRIBUTING.md for manual instructions. The Elastic documentation site exposes…
725mo agoReviewMountCreate a free account to mount
A React-based chat interface for Pydantic AI that uses Vercel AI SDK and Elements. The project consists of a frontend (Vite + React + TypeScript) and a Python backend (FastAPI + Pydantic AI). Note: Stop any logfire platform instances to avoid port 8000 conflicts. Set E2E_VIDEO=1 to record screen videos with slowMo. Set E2E_TEST_DIR=<path> to override which directory Playwright walks (defaults to tests/e2e; pnpm test:e2e narrows it to tests/e2e/deterministic). The offline suite ignores E2E_TEST_DIR — it runs off its own…
713mo agoReviewMountCreate a free account to mount
All deterministic tests drive the real FastAPI server at tests/server/server.py, which uses pydantic-ai's FunctionModel to return predictable streams. The models mapping is the authoritative fixture registry. A spec picks an entry via sendMessage(page, '<name>', '<message>'). When you need new behavior, add a FunctionModel to the server's models dict and select it from the spec. Do not introduce page.route mocks — they duplicated the SSE wire format and drifted on the SDK v5→v6 bump, which is why we switched.
713mo agoReviewMountCreate a free account to mount
For expensive build commands, ALWAYS use the run_task MCP tool instead of Bash. **Commands that MUST use run_task:** - gradle, bazel, make, cmake, mvn, cargo build, go build - docker build, docker-compose, kubectl, helm - npm/yarn/pnpm build, pytest, jest, mocha Usage: - command: The full shell command - working_directory: Absolute path to project root - env_vars: Optional like "KEY=value,KEY2=value2" NEVER run these via Bash. Always use run_task MCP tool.
6810mo agoReviewMountCreate a free account to mount
A file Claude Code reads at the start of every session. It holds the commands, conventions and warnings the agent needs for this project.
Where does it go?
At the repository root. Claude Code also reads CLAUDE.md files in subdirectories when it works there.
What should it contain?
Build and test commands, the project's layout, conventions that aren't obvious from the code, and mistakes to avoid. Short files tend to work better than long ones.
CLAUDE.md or AGENTS.md?
Claude Code reads CLAUDE.md; most other agents read AGENTS.md. Many projects keep one and point the other at it.
NVIDIAReads credentialsCLAUDE.md
