carbon-steel-pans/ is a real report produced by this system, not a
synthetic fixture. It shows the whole pipeline in one place:
| File | What it is |
|---|---|
source.md |
The research markdown the deep-dive skill produced — ~7,800 words, 9 subtopics, 90+ sources consulted |
report.html |
The published field packet the publish-report workflow built and polished from it — 120KB, self-contained |
meta.json |
The sidecar the library index reads |
Open report.html in a browser to see the house visual standard: stamped hero,
KPI stamp grid, ticket metadata, numbered sticky TOC, comparison matrices,
estimate-ticket figures with dotted leaders, callouts, and the source audit
trail — responsive down to 360px, with print CSS and reduced-motion handling.
library/ holds two real reports laid out as a ready-made library — the carbon
steel buying guide and a deep dive on getting AI coding agents to produce
non-generic web design (95 sources). It is what the README previews are
rendered from, and the two show deliberately different report shapes: a
product-comparison packet and a technical field guide.
fieldpacket init
cp -R examples/library/reports examples/library/meta "$(fieldpacket root)/"
fieldpacket index
fieldpacket servesource.md is a good reference for what the deep-dive skill should produce:
a research-date header, an executive summary that stands alone, thematic
sections (not concatenated subtopic notes), comparison tables, an uncertainty
ledger where sources disagree, and an annotated source list.