| name | sysml-new-project |
| description | Scaffold a new SysML v2 project under sysml-v2-models/projects/: folder layout, config.yaml, model files (connections, requirements, deploy, behaviour, root), READMEs, and repo index updates. Triggers: new project, create project, add project, scaffold project, new sysml project, start a project folder. |
| metadata | {"pattern":"pipeline","domain":"sysml-v2","version":"1.6","product":"memnet-llm==0.19.3","pairs_with":["sysml-root-config","sysml-requirements-generator","sysml-modeling-session-checklist","sysml-memnet-cache","sysml-memnet-documentation","mcp-memnet","sysml-v2-lsp-mcp","project-planner"]} |
| token_guardrails | - Ask for project slug, one-line purpose, and requirement ID prefix before bulk generation unless user gave them.
- Use an existing project config.yaml as OMG Kernel template; do not invent library paths.
- After scaffold: validate project load; update indexes; commit/push only when user asks.
- When MemNet is up at scaffold: AGENT-CONTEXT stub + catalog session skeleton (steps 11-12). Nested Snap: memnet-nested-sessions.
|
SysML new project
When: Greenfield sysml-v2-models/projects/<slug>/ — not fixing load order only (sysml-import-order-helper) or root/config drift only (sysml-root-config).
Pairing: In weft.Projects, hub .cursor/skills/sysml-v2-modeling + this skill for scaffold; elsewhere sysml-modeling-workflow + this skill. Use one specialist after (e.g. sysml-requirements-generator, sysml-connections) for content depth.
Before you scaffold
Confirm with the user (or infer from the request):
| field | rule |
|---|
slug | kebab-case folder e.g. delta-datacentre-heat-exchange-unit-test-bench |
purpose | One sentence for README and config comment |
req_id_prefix | Project-specific e.g. DDCHXU-R1 not generic R1 unless agreed |
package_prefix | Short PascalCase e.g. DeltaDCHXU for DeltaDCHXUConnections etc. |
libs_common | Only if deploy uses NI catalog or shared ports sbrio-9651-carrier-board; else Kernel+ISQ/SI temperature-iv-curve |
hardware | Optional hardware/<slug>/ via repo hardware-custom-pcba-workflow when PCBA in scope (weft.Projects) |
| For ambiguous scope or a roadmap, run project-planner first or state planning skipped per sysml-modeling-session-checklist. | |
Pipeline
Copy this checklist and track progress:
- [ ] 1. Create project folder + models/ + outputs/
- [ ] 2. config.yaml (OMG chain + project model_files, root last)
- [ ] 3. connections-*.sysml (if deploy has links)
- [ ] 4. requirements-*.sysml (optional but typical)
- [ ] 5. deploy-*.sysml (parts + system composite)
- [ ] 6. behaviour-*.sysml (optional)
- [ ] 7. root-*.sysml (imports only)
- [ ] 8. Project README + outputs/README.md
- [ ] 9. Repo indexes (projects/README, root README, docs/DOCS_INDEX)
- [ ] 10. Validate / visualize smoke test
- [ ] 11. AGENT-CONTEXT.md (thin stub — when MemNet or multi-session design expected)
- [ ] 12. MemNet catalog `session_open` + campaign `TSK` + `MOD` rows (when serve up). Prefer `snap_model` if a nest applies; do not flatten the tree into one session.
1. Folder layout
sysml-v2-models/projects/<slug>/
├── config.yaml
├── README.md
├── models/
│ ├── connections-<slug>.sysml # if needed
│ ├── requirements-<slug>.sysml
│ ├── deploy-<slug>.sysml
│ ├── behaviour-<slug>.sysml # optional
│ └── root-<slug>.sysml # always last in config
└── outputs/
└── README.md
File names: lowercase, hyphens; prefix matches slug: deploy-<slug>.sysml.
2. config.yaml
- Copy
model_files OMG block from temperature-iv-curve or delta-datacentre-heat-exchange-unit-test-bench config.yaml.
- Set top comment to project purpose.
- List project files in dependency order;
root-<slug>.sysml last (load-order).
Typical order: connections → requirements → deploy → behaviour → root.
3–7. Model files
| file | skill |
|---|
connections-* | sysml-connections / sysml-software-port-generator / sysml-physical-port-generator |
requirements-* | sysml-requirements-generator |
deploy-* | sysml-hardware-part-generator / sysml-software-part-generator / sysml-nested-structure-modeling |
behaviour-* | sysml-behaviour-generator |
root-* | sysml-root-config stub assets/root-package-stub.sysml |
| Root package imports all project packages; no deploy logic in root. | |
satisfy: Add on system part def only after requirement defs exist and names are agreed.
8. READMEs
- Project README: purpose, model file table (file → package), validate commands, req ID table if any.
- outputs/README.md: model-first note,
visualize.py --project <slug> example, pointer to sysml-view-doc-sync.
Do not put project docs under repo docs/ when the workspace defines sysml-docs-outputs (e.g. weft.Projects .cursor/rules/sysml-docs-outputs.mdc).
9. Repo indexes
Add one row each to:
sysml-v2-models/projects/README.md- Repo root
README.md (projects table + optional “Key docs” subsection)
docs/DOCS_INDEX.md (sysml-v2-models project list sentence)
Match existing table style; link to projects/<slug>/README.md.
10. Verify
From sysml-v2-models/ (venv + OMG submodule):
python scripts/visualize.py --project <slug> --diagram bdd --format svg
Or sysml-v2-lsp-mcp / mcp-sysml-v2 validate on edited files (repo may use sysml-v2-lsp-mcp under .cursor/skills/).
11. AGENT-CONTEXT.md (required when MemNet-assisted or multi-session design)
Create AGENT-CONTEXT.md at project root per sysml-memnet-snap.md. Max 40 lines:
# Agent context — <slug>
**MemNet session:** `<mn_…>` · **Anchor:** `TSK_model_<short>`
## Summary
<purpose one sentence + 5–10 line human overview as design grows>
## MemNet
Query `TSK_model_<short>` — do not duplicate topology/backlog here.
Derive <short> from slug (e.g. vedan-foam-detection-lite-ver2 → vfdl2). Record session id after step 12.
12. MemNet skeleton (required when MemNet is up at scaffold)
serve_status -- TCP/HTTP / unsure. If false, skip; complete step 11 with placeholder session. Multitask MUST NOT in-process.session_open with coding/SysML kinds enabled (see sysml-memnet-patterns.md for field notes). This is the catalog session.mutate skeleton (openCypher-shaped; leftover add / id:'NEW' named leftover). Campaign goal = TSK_model_<short> from the slug. If the load tree is already nested: snap_model per sysml-memnet-snap.md -- MUST NOT dump every file into this one session.
CREATE (t:TSK {goal: $campaign, phase: 'model', status: 'in_progress', recycle: 'persistent'})
CREATE (m1:MOD {path: 'models/deploy-<slug>.sysml', role: 'deploy', status: 'active', recycle: 'persistent'})
CREATE (m2:MOD {path: 'models/root-<slug>.sysml', role: 'root', status: 'active', recycle: 'persistent'})
MATCH (t:TSK {goal: $campaign}), (m1:MOD {path: 'models/deploy-<slug>.sysml'})
CREATE (t)-[:OWNS {note: 'scope', recycle: 'persistent'}]->(m1)
Add MOD for each other models/*.sysml created. Store returned catalog session_id in AGENT-CONTEXT.md and optionally MEMNET_SESSION.
Optional follow-ups
| need | skill |
|---|
system design report outputs/ | system-design-report-generator / sysml-view-doc-sync |
custom PCBA hardware/ | hardware-custom-pcba-workflow (weft.Projects repo skill) |
shared part libs/common/ | sysml-common-lib-contribution |
Rename later
Use git mv for folder and *-*.sysml files; update config.yaml, README paths, and indexes in one commit.
Reference
