By 3shn
AI-native SysML v2 systems-engineering co-pilot. Authors valid textual SysML v2 (.sysml), grounds every type/unit in the standard library (ISQ, SI, …), and validates with the real SysML v2 Pilot compiler in a self-correcting loop. Bundles an MCP server (validate_sysml_file, query_library, get_library_element).
An AI-native SysML v2 co-pilot. It authors valid textual SysML v2 (.sysml), grounds every
type/unit in the standard library (ISQ, SI, …) instead of inventing them, and validates with the
real SysML v2 Pilot compiler in a self-correcting loop so the output actually compiles.
sysml/
├── .claude-plugin/plugin.json # plugin manifest (auto-loads as `sysml@skills-dir`)
├── .mcp.json # declares the `sysml` MCP server
├── skills/sysml-copilot/ # the co-pilot skill (SKILL.md + references + examples)
└── mcp-server/ # the MCP server (validate + library introspection)
├── server.py # stdio MCP server (run via `python3`)
├── library_index.py # name→element index over the standard library
├── java/SysmlValidatorServer.java # warm Pilot-kernel validator (stdin/stdout JSON)
└── setup.sh # downloads the kernel jar + compiles the validator
validate_sysml_file(content | path, context_paths?) → {ok, diagnostics:[{line, column, severity, code, syntax, message}]}query_library(query, limit?) → ranked standard-library elements with qualified namesget_library_element(qualified_name) → an element's declaration excerptPATH. The MCP server is stdlib-only — no Python
dependencies to install, no venv, no uv.Systems-Modeling/SysML-v2-Release).setup.sh).# one-time: fetch the kernel jar and compile the validator
mcp-server/setup.sh
Configure paths via env if you need to override the defaults (which land in plugin-local .runtime/):
SYSML_LIBRARY_PATH=${CLAUDE_PLUGIN_ROOT}/mcp-server/.runtime/sysml.library
SYSML_KERNEL_JAR=${CLAUDE_PLUGIN_ROOT}/mcp-server/.runtime/jupyter-sysml-kernel-0.59.0-all.jar
.mcp.json)The MCP server is stdlib-only and launched directly by a system python3 (no venv, no uv — those
caused handshake timeouts -32000 when deps were resolved at launch). The .mcp.json uses
${CLAUDE_PLUGIN_ROOT}-relative paths so the plugin is portable and works immediately after
running setup.sh.
Run mcp-server/setup.sh once (downloads the kernel jar if missing, compiles the Java validator),
launch Claude Code from the repo root, then /reload-plugins and verify /plugin shows sysml
connected. The Java validator also auto-compiles on first use if setup.sh wasn't run.
Admin access level
Server config contains admin-level keywords
Own this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimOwn this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimBased on adoption, maintenance, documentation, and repository signals. Not a security audit or endorsement.
npx claudepluginhub 3shn/skill-sysml --plugin sysml-copilotA Julia-first Bayesian-workflow copilot: a Socratic super-REPL for building probabilistic models incrementally with fail-closed calibration gates. Tool-agnostic methodology in prose + math (no baked code); consult-the-docs discipline; drives a live Julia session via Revise.jl hot-reload and an MCP REPL.
Harness-native ECC operator layer - 67 agents, 271 skills, 92 legacy command shims, reusable hooks, rules, selective install profiles, and production-ready workflows for Claude Code, Codex, OpenCode, Cursor, and related agent harnesses