Changelog

All notable changes to this project are documented in this file. The project follows Semantic Versioning.

Unreleased

Added

• Added an Agent Plugins 1.0.0 root manifest so the canonical collection can be discovered as one portable plugin while retaining the existing Claude Code and Codex compatibility manifests and Skills CLI distribution path.

Fixed

• Aligned skills.sh section titles and descriptions with the portfolio catalog, and extended the freshness check to detect a stale flat repository page even when its skill inventory has partially updated.

Security

• Added narrow trust boundaries and adversarial eval coverage to feature-spike, bootstrap-repository, and review-mintlify-docs without removing their required user, repository, API, or documentation inputs.
• Added a pass-by-default daily skills.sh audit monitor across Agent Trust Hub, Socket, and Snyk. The four current medium Snyk W011 warnings are exact, owned, expiring exceptions tied to runtime safeguards and the upstream repository-content classification issue; new or changed findings fail.

Documentation

• Aligned the README's skill groups, ordering, and descriptions with the portfolio skills page and canonical skills.sh catalog.
• Recorded v3.2.0 archive integrity, exact-tag multi-client installation, downstream Agent Tooling adoption, and the remaining skills.sh page and snapshot lag.

[3.2.0] - 2026-08-02

Added

• Added integrate-app-intents for architecture-preserving Apple system integrations across Siri, Shortcuts, Spotlight, snippets, App Schemas, onscreen context, and AppIntentsTesting, with seven execution evals and twenty balanced routing cases.

Changed

• Added a non-media GRDB portability case and explicit mutation-idempotency guidance to integrate-app-intents, and removed project-specific provenance from the runtime guidance in tune-mobile-client-performance.

3.1.0 - 2026-08-02

Added

• Added tune-mobile-client-performance for evidence-led tuning of a specific SwiftUI or React Native interaction across motion, observation, rendering, media, network, and observability boundaries, with six execution evals and twenty balanced routing cases.

Documentation

• Separated general iOS capability decisions from mobile performance tuning: advanced pinned standalone XcodeBuildMCP into an agent-tooling pilot, kept Memgraph and ETTrace task-gated, and selected a first-party App Intents skill pilot from NextUp's existing shipped implementation rather than installing or forking build-ios-apps.

Fixed

• Made the skills.sh freshness monitor honor bounded Retry-After guidance and retry transient HTTP failures so catalog throttling is not mistaken for stale skill content.

Documentation

• Separated general iOS capability decisions from mobile performance tuning: advanced pinned standalone XcodeBuildMCP into an agent-tooling pilot, kept Memgraph and ETTrace task-gated, and selected a first-party App Intents skill pilot from NextUp's existing shipped implementation rather than installing or forking build-ios-apps.
• Recorded Patrick Skills v3.0.0 release and exact-tag installation evidence, downstream Patrick Plugins v3.4.0 and Agent Tooling v0.8.0 verification, the remaining skills.sh stale-snapshot gate, and the requirement to dogfood bootstrap-repository only on the next genuinely new or empty real repository.

3.0.0 - 2026-07-31

Added

• Added a local skills.sh refresh command and daily catalog-and-snapshot freshness check so the public repository page and stored skill files cannot silently drift from the canonical skill tree.
• Added bootstrap-repository for generator-backed Swift, Expo and React Native, Python, and Node repository setup at exploration, maintained, or shipping maturity. The validated skill includes a fixture-tested read-only local and GitHub inspector plus focused stack and policy references, without templates or remote mutation helpers.

Changed

• Updated skills.sh.json to the current documented grouping schema and aligned its design, product-development, and Mintlify sections with the README.

Removed

• Removed production-hardening after fresh, repository-backed comparisons found no reliable behavioral lift over a direct hardening request. Patrick explicitly rejected a compatibility-only deprecation window, so the next release must be a major version.

Documentation

• Added a concise audience and value statement to the README, removed the alternate marketplace-plugin installation note, and removed the unnecessary Node.js badge.
• Validated bootstrap-repository after four natural candidate-absent and four with-skill scaffolds confirmed proportional maturity selection, upstream generator provenance, one local quality contract, and an exact GitHub policy plan. Follow-up checks enforce one bounded generator-state retry and defer licenses and other publisher identity decisions; universal templates, reusable configuration fragments, and remote mutation helpers remain unnecessary.
• Recorded v2.4.0 publication, curated-note and archive integrity, exact-tag multi-client installation, Skills CLI discovery, Patrick Plugins v3.3.0, and Agent Tooling v0.7.0 evidence; marked the UI video and design-system candidates as released.
• Recorded the real-history, public-overlap, baseline, and released-skill audit that led to removing production-hardening instead of expanding it.

2.4.0 - 2026-07-31

Added

• Added audit-design-system-health for read-only repository-wide drift, consolidation, system-gap, and enforcement audits across native and web codebases.
• Added align-ui-to-design-system for bounded, evidence-backed conformance review and approved correction of a completed native or web UI surface.
• Added analyze-ui-video for evidence-first screen-recording diagnosis and reference reconstruction.
• Added a deterministic video probe and contact-sheet helper that distinguishes observed and nominal frame rates, plus focused tests, execution evals, and balanced routing cases.

Changed

• Grouped the README skill catalog into design and development skills and Mintlify documentation skills.
• Extended code-native-ui-ideation with an optional faithful, stretching, and divergent design-system distance axis grounded in protected foundations and intentional departures.

Removed

• Removed the obsolete mintlify-docs migration guide; its only user already migrated to this collection.

Documentation

• Added a mandatory skills.sh overlap gate before pilot, validation, and release, including source inspection, upstream/fork/compose decisions, provenance, and retirement criteria in each candidate brief.
• Recorded v2.3.0 publication, exact-tag installation, archive integrity, and marketplace evidence and marked feature-spike as released.
• Classified analyze-ui-video as one portable skill candidate for evidence-first screen-recording analysis, with reference-reconstruction and visual-bug branches grounded in three completed repository examples.
• Captured design-system-aware ideation as an enhancement to code-native-ui-ideation, plus separate candidates for component-level conformance and repository-wide design-system drift auditing.

2.3.0 - 2026-07-30

Added

• Added the feature-spike skill for bounded product-value and technical-feasibility experiments that end in a continue, change, stop, defer, or inconclusive decision.
• Added focused evidence guidance, a reusable feature-spike brief, five execution evals, and twenty balanced routing cases grounded in three completed proofs of concept and fresh no-skill baselines.

Documentation

• Recorded v2.2.0 publication and exact-tag installation evidence and marked the expanded feature-delivery skill as released.

2.2.0 - 2026-07-30

Changed

• Expanded feature-delivery from a thin implementation loop into an evidence-led coordinator for product and technical specification, cross-repository contracts, explicit execution-mode selection, integration, and staged user availability.
• Added a reusable feature-delivery template and focused references for specification, cross-repository compatibility, optional Superpowers or agent-team execution, and independently gated rollout and rollback.
• Expanded feature-delivery to six execution evals and twenty balanced routing cases, with baseline and fresh-context forward replays covering ambiguous architecture, mid-flight mobile/backend rollout, and proportional single-repository delivery.
• Replaced the old inventory naming and WF-* records with a skill candidate inventory and SC-* identifiers so the repository consistently describes the reusable artifacts as skills.

Documentation

• Recorded exact-tag installation evidence for v2.1.0 and marked the four migrated Mintlify skills as released.

2.1.0 - 2026-07-30

Added

• Consolidated and hardened the former mintlify-docs collection as scaffold-mintlify-site, review-mintlify-docs, generate-mintlify-reference, and write-mintlify-changelog.
• Added self-contained editorial references, standard assets, cross-runtime metadata, execution evals, routing evals, and clean-install coverage for the documentation skills.
• Added skill-local MIT metadata so licensing survives individual Skills CLI installations.
• Added a permanent migration guide from the retired mintlify-docs skill and plugin names.
• Added a security policy covering agent instructions, bundled scripts, and packaging risks.
• Added a release runbook for immutable tags, curated notes, exact-tag installation, and downstream marketplace sequencing.

Changed

• Replaced the former generic Mintlify skill names review-docs, document-reference, and changelog-writer with explicit globally installable names.
• Made the official Mintlify capability an optional source of current syntax and schema mechanics instead of a runtime-specific requirement for loading Patrick's editorial skills.
• Replaced the historical mirror-repository default with Mintlify's current direct repository and subdirectory model, while keeping all publishing actions explicit and separately authorized.
• Added deterministic contrast checking and hardened the bundled reference generators, starter MDX, favicon, and local validation commands.

Documentation

• Recorded the completed release, client-installation evidence, and lessons for the first audited design skill.
• Classified end-to-end feature development as a candidate expansion of feature-delivery, grounded in three private cross-repository feature histories and an explicit composition boundary with optional Superpowers and agent-team execution.
• Captured feature-spike as a separate evidence-seeking skill candidate whose outcome is a continue, change, or stop decision rather than production delivery.

2.0.0 - 2026-07-30

Changed

• Renamed the repository from pdugan20/patrick-workflows to pdugan20/skills so the public name matches the artifacts it distributes.
• Renamed the Claude Code and Codex plugin package from patrick-workflows to patrick-skills.
• Simplified the README around one Skills CLI install command and canonical repository links for each skill.

Migration

• Install directly with npx skills@latest add pdugan20/skills.
• Marketplace users should replace patrick-workflows@patrick-tools with patrick-skills@patrick-plugins.

1.1.0 - 2026-07-30

Added

• A researched skill authoring standard covering skill selection, portable structure, routing, evaluation, external tools, and review.
• A durable skill candidate inventory and evidence-based brief template for auditing Patrick's real design and development processes.
• Versioned execution and routing evals for code-native-ui-ideation, feature-delivery, and production-hardening, enforced by repository validation.
• An isolated Skills CLI installation smoke test for Claude Code, Codex, and Cursor, enforced in CI.
• A pinned external Agent Skills validation gate for structure, links, context size, orphaned resources, and content-quality heuristics.
• Platform-specific SwiftUI, React Native, and web guidance for code-native-ui-ideation, covering runnable variants with consistent fixtures and project-native development surfaces.
• Standard repository badges, direct skill source and skills.sh links, and installation paths organized by client and version behavior.
• Curated GitHub Release notes extracted from the matching changelog section.

Changed

• Expanded code-native-ui-ideation with an ideate-versus-build branch, a shared comparison contract, and an explicit graduation cleanup boundary.
• Replaced the claude-code-lint gate, dependency, and configuration with the exactly pinned official Claude Code plugin validator after a scope audit found duplicate, ineffective, and non-portable checks.
• Documented collection-level semantic versioning with per-skill history in this central changelog instead of non-portable skill-local versions and changelogs.

1.0.0 - 2026-07-30

Added

• The code-native-ui-ideation, feature-delivery, and production-hardening skills.
• Claude Code and Codex plugin manifests backed by one canonical skill tree.
• Skills CLI metadata, repository validation, spelling checks, workflow-security analysis, scheduled link validation, and automated release checks.