System Design Architect
An Agent Skill that maintains one system model and produces synchronized views for design and review.
- Agent Skills
- System Design
- Python
Under the hood
Technical architecture
This is a skill and a set of local validation/rendering scripts. It is not a hosted application or a live infrastructure monitor.
Working method
Agent Skill instructions
Guides evidence gathering, boundary definition, current-state review, target design and verification.
Canonical model
system.json · JSON schema
Records the system elements and relationships shared by the visual views.
Validation
Python scripts
Validates the skill and the canonical model before reporting on the design.
Presentation
Self-contained HTML report
Renders synchronized current, target, transition, risk and verification views from the model.
Failure and fallback path
Keep gaps and assumptions visible in the model and review. Live operational adapters and an executable behavioral evaluation harness are roadmap items in the repository, not capabilities implied by the report.
One request, end to end
- Gather evidence and establish the purpose and boundary.
- Update and validate the canonical system.json model.
- Render related views, inspect risks and define verification checks.
Why it is built this way
- One canonical model reduces contradictory diagrams.
- Use software and general-system lenses where each fits.
- Recovery authority stays bounded rather than silently changing system policy.