soria-stack
by Soria-Inc
Data pipeline skills for Soria Analytics. Cognitive modes for upstream pipeline work (scrape → extract → value-map → publish) and for building dives (dbt marts + manifest + React component + DivesPage registration + rows in the shared verifications seed + methodology wired into the component). All skills drive the Soria platform through the mcp__soria__* MCP tool namespace. SQL and React dives are authored locally in frontend/src/dives/ and shipped via git push + PR — there is no soria CLI. Soria skills: /tools (verify MCP + local stack), /env (preflight; no isolated envs), /status (what exists), /plan (ETVLR orchestrator), /ingest (scrape + extract + publish), /map (value mapping), /parent-map (centralized parent company resolution), /dive (build a dive end-to-end), /preview (render a dive in chat), /verify (prove data correct), /dashboard-review (adversarial browser QA against dev.soriaanalytics.com), /diagnose (failure triage), /ticket (file structured tickets mid-session), /promote (warehouse_diff + warehouse_promote + PR + CI), /newsroom (news pipeline ops), /lessons (retrospective), /browse (persistent headless Chromium for verification), plus engineering skills /dev-env, /dev-dives, /test, and /code-review for branch-local app work, dev HTTPS dive frontend repair, test evidence, and Soria-specific code review. Suggest the right skill by stage: starting a session → /tools; investigating what exists → /status; planning work → /plan; building a pipeline → /ingest; normalizing values → /map; resolving parent companies → /parent-map; building a dive or reviewing its SQL → /dive; proving data correct → /verify; testing live dive UI → /dashboard-review; something broke → /diagnose; filing a bug/feature ticket → /ticket; promoting to prod → /promote; news pipeline → /newsroom; reviewing recent work → /lessons.