Claude Skills by jLAM-ERR

Design how corp-llm-gateway varies detection/policy by country, division, or regulatory regime — reusing the profiles/ bundles, extensions/ registry, and DETECTOR_REGISTRY seams — without forking core or executing untrusted code. Produces an extension-point map, options, a recommended hybrid, and a worked example. Use for "plugin architecture", "add a jurisdiction/country/division/regime profile", "how do we extend for X".

Audit corp-llm-gateway's operational/product readiness — config resolution, management CLI, Helm/Docker deploy, observability, docs, packaging — and produce a prioritized gap report with a top-5 build list. Use for "readiness audit", "is this deployable", "what's missing to ship", "productionization gaps". Not for code correctness (use security-audit for leaks).

Whole-repo leak-surface security audit of corp-llm-gateway against its zero-leak criterion and the 6 CLAUDE.md invariants — sanitizer coverage, NEVER-gate, auth/tokens, placeholder bijection, egress/DLP. Produces CONFIRMED/SUSPECTED findings with file:line + fixes. Use for "security audit", "sec-audit", "find leak paths", "do the invariants still hold". For reviewing a pending diff instead, use the built-in security-review.