axis-closure-matrixlisted

# Axis Closure Matrix ## Purpose EITR's architecture has more surface area than can be audited all at once, and open-ended "let's review the whole thing again" passes never converge — any sufficiently complex codebase has near-infinite room for "could be more complete," so a fully open audit just burns tokens without ever reaching a release-ready verdict. This skill exists to replace that loop with a **bounded** per-axis process: pick one architecture area, research what a genuinely solid version of it needs from real external sources (never from memory), compare it line-by-line against what EITR currently does, and stop at an explicit, checkable Definition of Done — not at "looks good enough." An axis qualifies for a single pass of this skill only if it maps to **at most 3 independently-versioned subsystems** (e.g. "CI/CD generator" = 4 CI providers sharing one `cicd.ts` source = 1 subsystem; "AI-agent layer" = agents + skills + MCP bridge + slash-commands = 4 independently-versioned subsystems, so it does NOT qualify as one axis) **and** is expected to produce **at most ~15 Core+Extended table rows**. Examples that qualify as one axis: "the TS+Playwright CPOM base," "the CI/CD generator," "the Python CPOM base." If the named target exceeds either bound — "the AI-agent layer," "the entire codebase," or anything you cannot map to a single noun phrase after 1-2 clarifying questions — do not run the full process against it as one axis; instead propose 2-4 named sub-axes (e.g