skills

Use when reviewing an implementation plan, codebase, feature area, or pull request for likely defects, missed requirements, weak tests, security gaps, data risks, UX mismatches, or reliability issues before or after implementation.

Use when editing or revising a skill (SKILL.md), prompt, rule, system message, or instruction to trim bloat.

Use when writing, reviewing, or refining Gherkin, BDD scenarios, acceptance criteria, feature files, behavior specs, examples, user journeys, business rules, edge cases, or Given/When/Then wording.

Use when writing, reviewing, editing, or renaming tests, variables, functions, classes, types, components, scenarios, tickets, branches, or other identifiers where names may be vague, implementation-focused, misleading, or hard to understand.

Use when reviewing the active or named branch against a target branch, drafting GitHub review comments, verifying findings locally, or coordinating focused reviewer subagents.

Use for software development feature or bugfix work that should follow a strict red, green, refactor loop with isolated subagents, verified failing tests, minimal implementation, review gates, commits, pull requests, and ticket observations.

Use when validating that an Epic, Feature, parent ticket, or ticket subtree is coherent, dependency-safe, testable, and ready before implementation or TDD orchestration.

Use when building, grooming, or refining an Epic, Feature, parent ticket, or ticket tree before implementation so work can be manually reviewed, promoted from triage, and later executed with TDD.

Use when Codex needs to run an intense evidence-first pull request review against dev, develop, or a named target branch; gather requirements from GitHub PR text, Linear issues, branch names, tickets, and code; use vette risk lanes to find defects; create and verify temporary TDD-style failing tests for high-confidence failures; and draft PR review comments that include the generated test code.

Use when a parent ticket already has child tickets and a Codex/cmux orchestrator must sequence implementation PRs across ticket systems, branches, blockers, and worker tabs.

Orchestrate codebase research, decision clarification, and ticket decomposition into independently testable leaf tickets with plain-English acceptance tests.

Use when Codex needs to inspect, create, update, comment on, or otherwise manage Linear issues, teams, projects, milestones, initiatives, documents, labels, or Linear GraphQL data through schpet/linear-cli from the command line. Trigger for Linear CLI workflows, agent-friendly Linear issue queries, branch-aware issue lookup, creating GitHub PRs from Linear issues, or using `linear api` as a fallback for unsupported Linear operations.

Use when creating, editing, or triaging tracker-neutral tickets, issues, bugs, spikes, epics, features, or implementation tasks.

Investigate the existing codebase for behavior, reuse paths, dependencies, risks, and test coverage that matter to planning and ticket design.

Split a scoped outcome into independently testable tickets with clear dependencies, useful code references, and plain-English acceptance tests.