AI Agent Workflow Collection

Consult a stronger model mid-task for a second opinion before committing to an approach, when stuck, or before declaring a non-trivial task done — the fast-executor + strong-advisor pattern, implemented with the Agent and Workflow tools (Claude Code does not expose Anthropic's raw API advisor tool). Use when running on a cheaper/faster model for mechanical work and a hard design/architecture/risk decision needs a stronger check, or when authoring a Workflow script that should get expert review partway through instead of only at the end.

Use when stuck, wanting a second implementation or diagnosis pass, or handing a substantial task to Google's Antigravity CLI (agy) through a real subprocess -- a genuinely different agent harness, not just another model. Not relevant when already running under Antigravity CLI itself.

Delegate a substantial task to Claude CLI, optionally in the background, and track the resulting Claude job.

Use right after being corrected on something a CLAUDE.md file should already have told you, when explicitly asked to review CLAUDE.md, or periodically after a session with a lot of back-and-forth — reviews THIS conversation for friction and proposes a bounded CLAUDE.md edit. A lightweight, single-session complement to skill-tuning-loop (which mines many sessions and validates a proposal against held-out cases before surfacing it — reach for that instead when the ask is "make this systematic," not "check right now"). Own version of the dx plugin's /dx:review-claudemd.

Generate a solution, evaluate it against explicit pass/fail criteria, refine using the feedback, repeat until it passes or a round cap is hit. Use for tasks with a clear, checkable bar and demonstrable value from iteration — a bug fix that must pass a test suite, code that must satisfy a lint/style rubric, a document that must meet stated requirements. Distinct from advisor-checkpoint (one strategic consult before committing) — this is a convergence loop against a concrete standard.

Use at the very start of a new product idea — before a PRD, PCD, repo, or name exists, and any time someone says "I have an idea for", "should we build", "let's scope out", or asks for a PRD/PCD from scratch. Separates what is known from what is assumed, and forces the load-bearing assumptions to be measured before they become architecture. Works in Claude Desktop (research only) and Claude Code (research plus probes).

Use when PR checks fail, CI is red, or GitHub Actions workflows break - systematically inspects failing checks via gh CLI, pulls logs, checks for flakiness and scopes the breaking commit, scopes external checks, then creates fix plan using existing plan…

Use when asked to take a GitHub issue number, implement a fix, run builds/tests, commit with a closing message, and push. End-to-end workflow using gh CLI, local code changes, and git.

Use when touching widgets, animations, networking configuration, or background/observer cleanup — battery and performance patterns for WidgetKit timelines, SwiftUI animations, Low Power Mode handling, URLSession config, and NotificationCenter/Task cleanup under @Observable.

Use when wiring a new app-runtime service key (RevenueCat, Aptabase, ...) into a project — setting up Secrets.xcconfig, surfacing a key through project.yml into Info.plist, and reading it at runtime. Distinct from CI/server secrets, which live in GitHub Actions secrets and never touch an xcconfig — see the project's CLAUDE.md Secrets section for that split.

Copy-in CI recipes for adding macOS build/test/lint to a project on the ios profile — either a macOS-only app with no iOS target, or a hybrid app that also ships a separate macOS scheme sharing code with an iOS target. Use when scaffolding CI for a new macOS app, or adding a macOS target/scheme to an existing iOS project. The lacquer does not auto-sync macOS CI (only two fleet projects have needed it so far — not enough signal to justify conditional asset sync yet); these are reference recipes you copy into your project's own workflow files, adapted from two real, working implementations (a macOS-only app and an iOS+macOS hybrid).

Run a large batch of independent work items (a fleet-wide harvest, a multi-repo sync-down, an overnight backlog) as a persistent coordinator loop instead of one item at a time. Use when there are enough independent units of work that dispatch-review-merge would otherwise repeat many times in a row, or when the batch should keep making progress across a heartbeat interval instead of stalling until the next message.

On-screen agent alert: topmost message card + pulsating screen borders via Nameplate. Use before password-manager auth prompts or whenever blocked on the human.

Profile native macOS/iOS apps for CPU hotspots, hangs, and hitches using Instruments traces. Use when asked to identify performance hotspots, profile CPU usage, or diagnose slow code paths without opening Instruments.

Use when you need to find code, a pattern, or a past implementation across GitHub repos — including private ones you have access to — not just the current working directory. grep/Glob only reach the checked-out repo; this reaches everywhere gh is authenticated for. Own version of the dx plugin's /dx:private-github-search.

Create user-facing App Store release notes from git history. Use when asked to generate release changelog, App Store "What's New" text, or release notes based on git tags.

Scaffold, build, and package SwiftPM-based macOS apps without Xcode project. Use when you need a from-scratch macOS app layout, SwiftPM targets/resources, custom .app bundle assembly, or signing/notarization steps outside Xcode.

The bar a SKILL.md must clear before it ships in this lacquer — tight trigger-oriented frontmatter, single responsibility, no padding, companion files only when genuinely needed. Use when writing a new skill, reviewing an existing one, or auditing the skill set for quality.

Empirically test whether a lacquer skill needs an edit by mining real session transcripts for friction, proposing a bounded fix, and validating it against held-out cases before it ships. Use when a skill in core/skills/ or profiles//skills/ keeps getting corrected in practice, when triaging a scheduled tuning-pass PR, or when skill-authoring-standard's manual rubric isn't enough to tell whether a proposed edit actually helps. Distinct from skill-authoring-standard (a static prose rubric checked by a human) — this is an empirical, evidence-gated loop checked by rollouts.

Use when a test needs to wait for an async state change instead of a fixed delay — the notasksleepintests lint rule bans arbitrary Task.sleep in tests because it causes flaky failures. Provides the waitUntil polling helper to add to a test target instead.

Use when validating a user-provided or externally-sourced URL before it reaches AVPlayer, URLSession, or a WKWebView — building a positive- allowlist URL validator, or reviewing existing networking/media code for missing URL validation.

Use when building or reviewing a watchOS app or WatchKit extension — app structure and independent-app configuration, Watch Connectivity / companion-app sync, complications and Smart Stack widgets, controls or Live Activities on watch, background refresh and networking limits, watchOS-specific SwiftUI design conventions, or modernizing an older WatchKit/ClockKit project.

Use when working in a lacquer-managed project (one with a .lacquer.toml) and anything about that management comes up: the "No lacquer drift" CI job fails, lacquer audit exits 3/4/6, a lint config or CI workflow you edited reverts on the next sync, a hook or workflow needs changing, the project can't meet a baseline yet, a new stack (Swift package, backend, web app) is added to the repo, or a project needs onboarding.

Symbolicate and triage iOS/macOS crash reports (.ips, MetricKit, legacy .crash, Xcode Organizer .xccrashpoint) via the xcsym CLI. Use when given a crash file to diagnose, when TestFlight/App Store Connect crashes need triage, when a crash comes back unsymbolicated, or when asked what kind of crash something is.