Reading your first artifact bundle
What's in the ZIP, and how to validate it.
The artifact bundle is the deliverable. It's a ZIP file containing the generated systems-engineering documents, the source graph, and a manifest. Here's how to read it.
Bundle layout
bundle.zip
├── manifest.json # Index of every artifact in the bundle
├── stakeholder/ # NGOs, MOEs, operational scenarios
│ ├── ngos.json
│ ├── moes.json
│ └── scenarios.json
├── requirements/ # All requirement nodes
│ ├── functional.json
│ ├── performance.json
│ ├── interface.json
│ └── constraint.json
├── decomposition/ # Functions, components, interfaces
│ ├── functions.json
│ ├── components.json
│ ├── interfaces.json
│ └── dtrs.json # Design-to-requirement allocations
├── design/ # Per-component design specs
│ └── *.json
├── trade-studies/ # Decision analysis
│ └── *.json
├── risks/ # Risk register
│ └── risks.json
├── gates/ # Phase-gate readiness
│ └── phase_gates.json
└── render/ # Markdown render of each section
└── *.md
Where to start reading
For a quick scan: open render/index.md — it's a human-readable
walk-through linking to every section.
For a structured read in order:
- Stakeholder NGOs — confirm the AI captured the right stakeholders and their needs
- MOEs — measures of effectiveness. If these are wrong, the design downstream will be optimizing for the wrong thing
- Functional requirements — what the system must do
- Components — how the system is decomposed
- Trade studies — where the design picked one alternative over another, and why
- Risk register — what could go wrong, ranked by likelihood and impact
- Phase-gate assessments — readiness for SRR / SDR / PDR / CDR
Verifying the output
Start with citations. Each artifact node has a citations field
pointing at the section of your Brief it came from. Three checks:
- Citation precision: does the cited Brief passage actually support the artifact claim? The pipeline's verifier checks this, but you should spot-check 5–10 artifacts at random
- Coverage: did the pipeline miss anything important from your Brief? Compare the stakeholder list to your Brief's narrative
- Plausibility: does the design pass a senior-engineer sniff test? If something feels wrong, it probably is
In-browser explorer
You don't have to download the ZIP. From the run's Complete page, click Explore artifacts to browse the bundle in a tree view with side-by-side citation panel.
What to do next
If the run looks good: download the bundle, archive it. The Brief + bundle pair is your full audit trail.
If something needs adjustment: use Revisions to rerun a specific node with feedback, instead of rerunning the whole pipeline.