gestrich

Manages ClaudeChain automated PR chains. Use when working with claude-chain projects, listing open chain PRs, rebasing chain branches, triggering chain workflow runs, checking chain capacity, fetching workflow logs, creating new chains, or updating/editing an open chain PR locally.

Debugging guide for AIDevTools (evals, plan runner, Mac app, PRRadar). Shows how to discover configured repositories, eval cases, artifact paths, plan storage, log files, and CLI commands for troubleshooting. Use this skill when: the user reports a bug, asks to check eval data, shares a screenshot of eval results or the Mac app, mentions a failing case, wants to inspect provider output, needs to debug plan generation or execution, or is debugging PRRadar behavior (pipeline output, rule evaluation, Mac app issues). Screenshots of the Mac app are a strong signal to invoke this skill — they mean the user wants you to reproduce or investigate using the CLI.

Reference for where Claude and Codex provider data is stored on disk — session logs, raw output, thinking/reasoning availability, and format differences between the two providers. Use whenever investigating provider output, debugging missing thinking blocks, reading session history, or asking "where does X get stored for Claude vs Codex?"

Running, writing, and extracting screenshots from UI tests and interactive tests for the AIDevTools Mac app. Use this skill when: running XCUITests, capturing app screenshots, adding new screenshot tests, extracting images from .xcresult bundles, or using the interactive test runner (XCUITestControl) to programmatically drive the app UI.

Guidance for this project's composition root pattern — how SharedCompositionRoot centralises shared service wiring, how the Mac app and CLI targets each have their own root that wraps it, and how platform-specific additions (models, git client, credentials) are layered on top. Use this when adding new shared services, creating a new CLI command that needs services, modifying how providers or credentials are wired, reviewing code that constructs services outside a composition root, or any time someone asks how dependency wiring works in this project.

Guide for adding, modifying, or reviewing configuration in this Swift app. Use when adding credentials or API keys, creating a new config file, adding a ServicePath case, setting up a service that reads settings, wiring configuration through the Apps layer, reviewing whether config is injected correctly (resolved values vs raw services), asking 'where should this config go?', or adding per-repo feature settings. Also use when working with SecureSettingsService, CredentialResolver, SettingsService, DataPathsService, or RepositoryConfiguration anywhere in the codebase.

Debugging context for PRRadar. This content has moved — use the ai-dev-tools-debug skill instead, which now covers PRRadar debugging inline.

Checks and fixes code quality issues in Swift: fallback values that hide failures, backwards-compatibility shims for unreleased code, raw String/[String:Any] types where typed models should exist, optional properties that should be non-optional, AI-changelog comments that describe what changed rather than why, duplicated logic across call sites, inline string formatting for derived identifiers (branch names, keys, paths), raw string literals used as shared identifiers, and force unwraps. Use when reviewing Swift code for quality, before committing, when asked to clean up or harden code, or when ai-dev-tools-enforce is running.

Structure CLI applications using command dispatcher pattern. Covers entry points, command routing, argument parsing, and explicit parameter flow. Use when building CLI tools or refactoring command structure.

Creates new Python service classes following Martin Fowler's Service Layer pattern with constructor-based dependency injection, proper layering (Core vs Composite), and static vs instance method conventions. Use when adding new services or refactoring business logic into service classes.

Determines the correct architectural layer (Entry Point/Service/Domain/Infrastructure) for code placement in Python applications using Service Layer pattern. Prevents layer violations and maintains separation of concerns. Use when adding new functionality, refactoring code, or unclear where code should live.

Writes unit tests for Python service classes using Arrange-Act-Assert pattern with proper mocking at boundaries. Tests behavior, not implementation. Mocks external systems only (API calls, file I/O, databases). Use when writing tests for services or fixing test coverage.

Follow Python code organization conventions including method ordering, datetime handling, circular import avoidance, and type annotations. Use when organizing service classes, handling dates, or structuring modules.

Apply constructor-based dependency injection with explicit configuration flow. Covers required dependencies, fail-fast principles, and avoiding optional parameters with default factories. Use when designing service constructors or managing dependencies.

Design domain models following parse-once principle with type-safe APIs, factory methods, and Repository pattern. Use when creating domain models, parsing data structures, or organizing business logic in models.

Analyzes git diffs and generates structured markdown summaries of code changes. Use when reviewing commits, understanding what changed in a branch, creating change documentation, or reviewing pull requests.

Debugging context for PRRadar with its configured repositories. Covers where rules, output, and settings live, and how to reproduce issues from the Mac app using CLI commands. Use this skill whenever debugging PRRadar behavior, investigating pipeline output, reproducing bug reports from the Mac app, or exploring rule evaluation results. Also use when Bill shares screenshots showing issues in the Mac app, or mentions code reviews, rule directories, output files, or PRRadar configs.

Create a new PRRadar review rule. Use when the user wants to add a code review rule, create a violation check, or set up a new pattern to detect in PRs. Handles choosing the project config, rule directory, evaluation mode (AI/script/regex), file targeting, and grep filtering.

Add an item to the project TODO list with t-shirt size categorization (Small, Medium, Large)

Verify recent PRRadar changes by running the CLI against the test repo

Creates automated UI test for a view, runs it, and captures screenshots to ~/Downloads. Use when the user asks to create screenshots, capture UI images, test view rendering, or generate visual documentation for a ViewController or SwiftUI view.

Interactively controls XCUITest through a CLI. Claude reads UI state and screenshots, decides actions, and executes commands via the CLI. Use for dynamic UI exploration, complex navigation flows, or when pre-scripted navigation isn't feasible.

Creates automated UI test for a view, runs it, and captures screenshots to ~/Downloads. Use when the user asks to create screenshots, capture UI images, test view rendering, or generate visual documentation for a view.

Interactively controls an app through XCUITest via a CLI. Claude reads UI state and screenshots, decides actions, and executes commands. Use for dynamic UI exploration, complex navigation flows, or when pre-scripted navigation isn't feasible.

Help the user manage their project — capture raw transcripts and brain dumps, show what's in motion and what's next, plan the coming sprint, and reconcile local plans against external sources of truth. Fire on (capture) "here's my brain dump", "voice transcript", "dump this into notes", or any long unstructured first-person paste that reads like a spoken transcript; (status) "what's the plan", "what's next", "catch me up", "where are we on X", "what's the status"; (sprint) "what should I do next", "plan my sprint", "what's on deck this week", "what should I focus on"; (reconciliation) "sync plans", "what's drifted", "what should graduate", "is the local plan still aligned with Jira/Confluence". Trigger even when the user does not name the mode — the skill picks the right mode from the ask and reads the project's AGENTS.md to figure out which planning sources apply (Jira, Confluence, a local plans/ folder, Slack-as-context, or some combination).

Onboard an existing project to the project-planning plugin's conventions — survey what's already there (notes, plans, AGENTS.md, commit history hints of Jira/Confluence/Slack usage), then propose and apply a structured plan of changes (folders to create, files to move, AGENTS.md content to add, planning sources to declare, plugin skills to enable). Use this skill whenever the user says "bootstrap this project", "set up project planning here", "make this project work with the planning plugin", "onboard this repo to project planning", "get this project ready for the planning skills", "wire this up to project planning", or anything along the lines of "land me in a new repo and make it usable by these skills". Trigger even when the user does not name the skill explicitly — if they're asking how to get a fresh project into shape so the other skills will work, this is the right home. Always shows the plan-of-changes for confirmation before writing anything; never silently edits.

Analyze recent project activity against the project's documentation and propose updates — clarifications, recurring questions that deserve a permanent home, decisions made in chat that the docs don't yet reflect, and gaps between code and the technical docs that describe it. Use this skill whenever the user says "update the docs", "what's missing from the docs", "review last week's discussion for doc updates", "what should we have written down", "are our docs in sync with the code", "any doc gaps from recent PRs", or otherwise asks for a documentation hygiene pass driven by what's been happening recently. Trigger even when the user does not name the skill — if they're asking what should be captured in the docs based on what's happened recently in chat or in the code, this is the right home. The skill reads the project's AGENTS.md to discover what activity sources to scan, where docs live, and how docs get updated; it then proposes updates the user can act on. Read-only by default; any write or hand-off only h

Provides the 4-layer Swift app architecture (Apps, Features, Services, SDKs) including layer responsibilities, dependency rules, placement guidance, feature creation, configuration, code style, and reference examples. Use when the user asks about the architecture, needs to understand layers, wants to add code, create features, or review architectural compliance.

Provides SwiftUI Model-View architecture patterns including enum-based state, model composition, dependency injection, view state vs model state, view identity, and observable model conventions. Use when building SwiftUI views, creating observable models, implementing state management, or connecting use cases to the UI.

Guides creation and editing of Claude skills. Use any time skills are being created, edited, or improved. Also use when understanding skill structure or learning best practices for skill authoring.

Next Phase Command

Plan Bill Command

Review Bill Command