skills

Use when Codex needs to review current Git changes, commit ranges, patches, or branch diffs as a Staff Software Engineer. The skill inspects diffs, tests, and relevant surrounding code; challenges design assumptions; focuses on high-impact production risks; and reports findings in a severity-ordered, staged format.

Use when Codex needs to analyze a Java project and create or update a Dockerfile at a requested path or under containers/ by default, using Oracle Java images, a JLink custom runtime for Java 9 and later, container-aware JVM memory settings, a non-root user, and docker build validation.

Use when Codex needs to create an opinionated Java Maven functional-test project for a web application, using JUnit 5, Playwright, Java 25, a disposable Docker container application harness, a browser harness, and small domain-specific fluent page objects derived from existing manual or automated functional behaviour.

Use when Codex needs to implement one existing implementation task from docs/tasks/. The skill verifies task readiness and dependencies, keeps work scoped to the selected task, prefers coarse behavioral or integration tests tied to acceptance criteria, applies a red-green-refactor loop where useful, designs for low cognitive load and deep modules, runs validation, and reports test coverage, manual checks, blockers, and follow-up work.

Use when Codex needs to convert vertical slices, a path-to-production plan, and repository context into ordered implementation task Markdown files. The skill derives small, independently grabbable tasks, preserves traceability to slices and production-readiness work, distinguishes product work from release-readiness work, records dependencies and readiness, and avoids publishing remote tracker issues unless explicitly asked.

Use when Codex needs to convert product requirements, vertical slices, and repository context into a production readiness and release sequencing plan. The skill identifies the first releasable increment, deployment target, environments, CI/CD, configuration, data, security, observability, rollback, validation gates, operational ownership, risks, and open questions without turning production planning into unrelated feature scope.

Use when Codex needs to convert a clarified Product Requirements Document into ordered vertical slice Markdown files. Prefer PRDs under docs/requirements/ as input; use scope documents only when the PRD step is intentionally skipped. The skill runs a slice readiness check, asks for clarification when slice boundaries or ordering are ambiguous, derives user-visible end-to-end slices, avoids horizontal technical task breakdowns, and writes slice documents under docs/slices/.

Use when Codex needs to convert an agreed product scope, MVP scope, or clarified feature scope into a Product Requirements Document before deriving vertical slices or implementation plans. The skill expands scope and lean-inception-style decisions into product-facing requirements, goals, non-goals, user journeys, acceptance criteria, success measures, constraints, and risks; runs a clarification loop for unresolved product decisions; and avoids writing a PRD that is not ready for slicing unless the user explicitly defers open questions.

Use when a user proposes a product idea, project, feature, architecture change, implementation plan, or other ambiguous change and needs help clarifying scope before acting. Clarify the goal, first user, first useful flow, constraints, trade-offs, risks, decisions, and open questions; inspect available local context first; run the clarification loop before writing scope; ask one decision at a time; write or update scope documents under docs/scopes/ only when enough clarity exists.

Use when Codex needs to identify, create, update, supersede, or review Architecture Decision Records (ADRs) for durable software architecture choices. This skill always writes concise Michael Nygard-style ADRs with Status, Context, Decision, and Consequences. Trigger this skill when work involves architecture decisions, technical trade-offs, platform choices, data ownership, integration patterns, security boundaries, deployment topology, persistence strategy, API contracts, cross-cutting constraints, or when existing planning artefacts mention ADR candidates or unresolved architecture decisions.