"Bridge is not reachable"
CATIA Magic must be open with a SysML v1 project loaded and the jgs-magic-sysmlv1-mcp plugin active. Check that first; it is almost always the answer.
From install to first insight.
Everything you need after the download: what must already be running, the one command that starts every task, the direct invocations for each specialist, and the two failures people hit first. The deep reference lives in skill-usage.md; this page is the path most people take.
§G1 · Install
One prerequisite decides everything: the jgs-magic-sysmlv1-mcp bridge must be installed and running, with a SysML v1 project open in CATIA Magic. The FREE read-only tier is enough for every skill in this pack. Then install the skills for your agent:
# preview first; install for Claude Code with the plain command python install.py --dry-run python install.py # another agent? ZCode installs the same way as Claude python install.py --agent zcode python install.py --list-agents # every target and path
Skills land flat under ~/.claude/skills/ (or your agent's equivalent) so short names like /jgs-v1 resolve. Restart your agent afterwards.
§G2 · Invoke
Start with the dispatcher and describe the task in plain English. It pings the bridge, checks your licence tier, and routes to the right specialist:
/jgs-v1 I just inherited this model: what am I looking at? /jgs-v1 Run a full health audit /jgs-v1 What would break if I change the DataBus block?
Each specialist also takes a direct invocation when you already know what you need:
/jgs-v1-navigate [<package-name>] structured model overview /jgs-v1-inspect <element-name> everything about one element /jgs-v1-impact <element-name> blast radius before a change /jgs-v1-audit [<root-package-id>] six-specialist health audit
Direct invocation skips the dispatcher's liveness check, so a down bridge shows up as raw tool errors instead of a friendly message. Anything that modifies the model (writes, v1→v2 migration, cross-model compare) is out of scope here: the dispatcher says so and offers the read-only equivalent.
§G3 · The specialists at a glance
| Skill | Reach for it when |
|---|---|
/jgs-v1 | you can name the goal but not the specialist; it routes and flags PRO-only requests |
/jgs-v1-navigate | the model is unfamiliar: packages, element counts, diagram inventory |
/jgs-v1-inspect | one element needs a full dossier: structure, ports, relationships, allocations |
/jgs-v1-search | you know part of a name and want ranked hits with type, qualified name, ID |
/jgs-v1-impact | a shared element is about to change and you need dependents first |
/jgs-v1-audit | you want the consolidated, severity-graded model-health audit |
/jgs-v1-report | the analysis has to be shared: model-health report plus RTM, written to a file |
/jgs-v1-fixplan | an audit is done: turn findings into a read-only remediation plan |
Plus the audit specialists (naming, docs, requirements, duplicates, unused, methodology), /jgs-v1-diagrams, /jgs-v1-units, /jgs-v1-status, and /jgs-v1-feedback for reporting gaps. All 18 are listed in SKILLS.md.
§G4 · When something looks off
Results seem truncated
Some bridge tools cap at 50 results or 2000 elements. Use qualified names for exact lookups, or scope the skill to a specific package.
Still stuck? The issue chooser routes bugs and improvements, and /jgs-v1-feedback drafts the report for you in-session. Full per-skill walkthroughs: skill-usage.md; other agents: other-agents.md.