Infrastructure & Observability Discovery

This skill guides the agent through a systematic discovery of a company's tech stack, observability stack, infrastructure scale, observability costs, and operational pain points. The output is one self-contained HTML report: an executive summary up top (scannable in under 2 minutes), collapsible per-tool deep dives below, and a provenance appendix linking every measured figure to the raw API response that produced it.

For install instructions, see README.md.

Division of labor (the core rule)

You (the agent)                      Collector scripts (deterministic)
────────────────────────             ──────────────────────────────────
detect environments & tools     →    collectors/<tool>/collect.py
ask qualitative questions       →    compute every volume/cost figure
write context.json              →    write summary.json + evidence/
run the report generator        →    report/generate_report.py writes the HTML

You never compute a reported figure. No improvised PromQL, no unit conversions, no arithmetic on API responses, no editing the report HTML. Every number in the report comes from a collector's summary.json; everything you contribute goes into context.json and is rendered as qualitative or "reported — not verified". If a collector fails, the answer is a documented gap — never a substituted estimate.