QA Skills — ISTQB skills pack for AI agents

The problem

Without explicit guidance, an AI agent "invents" test cases from generic intuition: random coverage instead of formal techniques. The full QA cycle — story verification, case design, execution, defects, CI/CD, continuous improvement — is left without a consistent framework the agent can follow.

The approach

QA Skills is a pack of 11 installable "agent skills" (npx skills add javalenciacai/QASkills) covering that full cycle, including case design with ISTQB techniques: equivalence partitioning, boundary value analysis, decision tables, state transitions. With the ISTQB design skill loaded, the agent applies the same formal techniques a senior human QA would use. An orchestrator (multi-agent-orchestration) decides which skill applies to each task; if no installed skill covers what's asked, it escalates to a discovery skill and, if that finds nothing either, to a skill that generates a new one — an attempt to close its own capability gaps at runtime, not just delegate to what already exists.

Result

MIT licensed. It's the repo with the most organic traction across the author's whole GitHub account: 5 stars and 3 forks. It installs into any project with a single command and gives an AI agent the same methodological rigor an ISTQB-trained QA would apply by hand — without the team having to document it from scratch.