codex-toolkit / skills

delegate-with-mission-cards: Model-agnostic Mission Control for bounded multi-agent work with portable reader/writer roles, explicit ownership, safe parallel waves, runtime-inherited model settings, and parent-side verification.
repository-intelligence: Evidence-backed architecture, ownership, dependency, hotspot, change-impact, and edit-conflict mapping.
multi-agent-work-coordinator: Safe work graphs for parallel tasks across code, documents, research, datasets, configuration, and other independently ownable resources.
codebase-evolution-controller: Dependency, framework, schema, and API evolution with compatibility stages, rollout, rollback, and evidence.
verification-and-release: Risk-based test strategy and evidence-backed release readiness decisions.
debugging-investigator: Reproduction-first debugging with ranked hypotheses, causal tracing, minimal fixes, and regression design.
documentation-synchronizer: Code-to-documentation drift detection and synchronized user, developer, API, configuration, migration, and operations updates.
product-design-director: Product UX and visual direction, brand extraction, responsive systems, accessibility, redesign audits, and taste calibration.
screenshot-to-interface: Maintainable reconstruction from screenshots with structural analysis, asset handling, responsive inference, accessibility, and visual fidelity loops.
production-web-builder: Production web architecture and implementation across data, forms, motion, accessibility, performance, testing, observability, SEO, and browser verification.
mobile-architecture-director: Evidence-based selection among Flutter, Expo/React Native, native, and other mobile approaches.
flutter-production-builder: Production Flutter architecture and implementation across state, routing, data, offline, platform integration, testing, performance, and release.
expo-react-native-builder: Production Expo/React Native architecture and implementation across routing, builds, updates, native modules, data, offline, testing, performance, and stores.
review-and-refactor-code: Evidence-backed change review and behavior-preserving refactoring with a mandatory proposal and approval gate.
optimize-codebase-performance: Measurement-first performance diagnosis and bounded optimization with comparable benchmarks and approval before edits.
codebase-improvement-planner: Repository-wide improvement discovery that classifies Major, Medium, and Minor opportunities, ranks them by evidence and value, recommends the best next upgrade, and stops for approval before broad edits.
typescript-quality-enforcer: Evidence-backed TypeScript/JavaScript quality audits, staged anti-slop Oxlint adoption, anti-laundering remediation, and permanent lint/type-safety enforcement.
content-provenance-hygiene: Evidence-first inspection and authorized sanitation of provenance and metadata surfaces in user-owned text, images, PDFs, and document containers through an optional local service, with approval gates and before/after verification.
unlazy: Evidence-backed completion discipline for substantial tasks using explicit gates, safe checks, natural-joint decomposition, final-candidate rechecks, and audited completion claims without overriding specialist safety or approval boundaries.
bug-finder: Evidence-backed hunting for previously unknown correctness defects using explicit invariants, high-risk surface prioritization, bounded proof/falsification, and clean handoff to causal debugging or the owning specialist.
toolchain-preflight: Narrow local execution-path resolution for unknown or broken shells, runtimes, package-manager shims, native quoting, browser capability, and text encoding before retrying builds, tests, scripts, or browser checks.
security-review: Focused security code review across trust boundaries, authorization, untrusted input, secrets, file/network handling, and business-logic abuse.