Skip to main content

​
The Story

Layer 3 gave STRATT the ability to compute blast radius. But the blast radius sat in a library — you had to write code to see it. Now it shows up on your PR as a Mermaid graph. Open a pull request that touches a prompt unit, and STRATT tells you exactly what breaks, what’s protected, and whether a human needs to approve before merge.

​
Key Messages

For the design journal (stratt.engineer): The STRATT CI workflow is now live — a reusable GitHub Actions pipeline (workflow_call) that any content repo can adopt in 10 lines of YAML. It runs all 8 failure modes in parallel across domain directories, posts a Mermaid dependency graph as a PR comment, and auto-labels PRs that touch gate or protected dependencies. The workflow consumes the same runCIPipeline() function that has 35 unit tests — no logic in shell scripts, no drift between CI and library. For developer audiences (Show HN / dev community): Your PR changes a shared prompt fragment. STRATT builds the dependency graph, computes which chains are affected, classifies each as gate, protected, or standard, renders a Mermaid flowchart, posts it as a PR comment, and blocks auto-merge if the blast radius hits anything sensitive. All 8 failure modes checked. Zero manual steps. The workflow is one uses: line in your repo. For LinkedIn / professional network: STRATT CI shipped today. When you change a prompt unit, the PR now shows a Mermaid dependency graph of every affected unit, colour-coded by risk. Gate dependencies block merge. Protected agents require approval. Contract breaks require major version bumps. 8 failure modes enforced automatically, computed by the same library that has 35 unit tests — not reimplemented in shell scripts. This is what prompt governance looks like at engineering maturity.

​
Content Opportunities

  1. Design journal entry: “The PR that shows its own blast radius” — screenshot of a PR comment with the Mermaid graph, walk through the 3-job workflow architecture
  2. Technical post: “Reusable CI for prompt architecture” — how workflow_call + domain matrix + comment-tag idempotency creates a plug-and-play governance pipeline
  3. Thread/short-form: Before/after — “Left: your prompt PR with no context. Right: the same PR with a Mermaid dependency graph showing exactly what breaks.”
  4. A2A ops angle: The entire workflow + 2 CLI commands + 2 library modules were designed and implemented in a single agent session by referencing the existing @stratt/graph library API. The agent read 15+ source files, identified that runCIPipeline() already did everything, and built thin wrappers rather than reimplementing.