Skip to main content

​
Summary

Baseline audit of five repositories that form the knowledge, testing, prompt engineering, and tooling backbone of the Devarno ecosystem. Purpose: establish measurable starting points for learning outcomes, agentic operation maturity, and marketing campaign coverage so future sessions can track delta.

​
Repositories baselined

​
so1-content (content hub)

38 findings documents, 9 blog posts, 4 social campaign directories, 30 agent documentation pages, and 14 operational runbooks. 27 files contain marketing or campaign keywords. This is the primary publishing pipeline for technical content and campaign material.

​
atlas (business knowledge base)

35 learning documents across Sparki, V01T, SO1, and family products. 695 total markdown files. 14 learnings reference agentic or autonomous operations. Marketing content lives in sparki/assets/campaigns/, v01t/marketing/, and monetisation strategy docs. Atlas is the richest single source of business context across the ecosystem.

​
veritas (prompt engineering)

298 prompts in the index (116 graduated, 182 unreviewed). The README states 255 — a discrepancy that needs reconciliation. 8 v3 agent suites with 62 XML specifications and 44 runbooks. 14 prompts are marketing-specific (campaign creation, social media, branding, SEO, landing pages). The FORGE pipeline prompts and A2A handoff chains represent the most mature agentic operations in the ecosystem.

​
ariel (testing framework)

4 baselines (2 active, 1 draft, 1 missing status field). 41 test definitions. 80% CI coverage gate. 3 files document learning outcomes. 18 agentic operation instances across 4 files. Zero marketing campaigns — ariel delegates all marketing to atlas by design.

​
traceo-mcp-server (MCP tooling + infrastructure)

19 production MCP tools (9 core, 10 Ariel integration). 154 test functions across 33 files. 13 Terraform configs, 28 Kubernetes manifests, 2 Helm charts, 22 CI workflows. No dedicated learning outcome documents or marketing campaigns — this repo is pure infrastructure.

​
Business impact

Learning outcomes are concentrated in atlas (35) and so1-content (38) with light coverage in ariel (3) and none in traceo-mcp. Future sessions that produce learnings should always commit to atlas; findings always to so1-content. Agentic operations are deepest in veritas (971 grep matches, FORGE pipeline, multi-agent chains) and growing in ariel (18 instances). The MCP server enables agentic operations but doesn’t document its own patterns — worth addressing as the tool count grows past 20. Marketing campaigns exist in atlas (22 files), so1-content (27 files), and veritas (14 prompts + reference template). Ariel and traceo-mcp have zero, which is correct for their role. The veritas/reference/marketing-campaign-base.md template should be the canonical starting point for new campaigns.

​
Action items

  • Reconcile veritas prompt count: index.json (298) vs README (255)
  • Add status field to ariel’s draft baseline
  • Consider adding a learning-outcomes doc pattern to traceo-mcp once tool count exceeds 20
  • Use this baseline to measure delta at next quarterly review