iql-context

Kind context (agentic-qa-core/references/skill-composition-strategy.md §2b). Shipped upstream and synced verbatim; a project adapts it through the two seams in §"How a project adapts it", never by editing this file.

Compact Rules

• DO: answer "why is the process shaped this way" from the index below, then point at the canonical reference for the procedure; never restate a DoD, a transition table or a link catalog that a reference already owns.
• DO: name stages by word (Shift-Left, Planning, Execution, Reporting, Documentation, Automation, Regression, Observation), never by number; a numbered "Stage N" in an older doc resolves through agentic-qa-core/references/stage-gates.md.
• DO: read .agents/project.yaml → qa.methodology before saying which stages a project runs or how strict its gates are; the shipped values are a default, not a decision.
• DO: read references/project-overrides.md for the project's own rules and exceptions; they win over this index for THAT project, and only there.
• DO NOT: resolve a web-vs-repo disagreement on your own. The repo's executable doctrine (agentic-qa-core/references/*) is what the skills enforce; the website is the public narrative. State both readings and cite both.
• DO NOT: treat this skill as SUT knowledge. Entities, endpoints, infra and Jira content live in .context/ maps and the project's <aspect>-context skills.
• DO NOT: edit this SKILL.md or its synced references in a consumer project; propose a refinement per agentic-qa-core/references/skill-refinement-protocol.md (project-owned prose goes in references/project-overrides.md).
• WHEN a question is about ONE aspect (phases and steps, the agentic contract, the eight approaches, the thirteen invariants): load only that reference from §"References".

Read full SKILL.md when: building a briefing that must explain the methodology to a subagent, answering "why" questions about the ladder or the stages, or checking which invariant a proposed shortcut violates.

What this skill knows