tyroneross

Compose a complete, durable build-loop handoff document from the current run state, and optionally launch a fresh session with it injected. Use when crossing a context boundary (context limit, planned restart, worktree GC). Triggers: 'hand off', 'handoff', 'new session', 'context limit', 'restart', 'fresh session', '/build-loop:compose-handoff'.

Use when build-loop needs UI design direction, visual style selection, UI guidance inventory, a .build-loop/app-contract/ui.md design contract, or non-trivial web/mobile/native UI planning. Build-loop-owned design route for design-contract-specialist; selects from project tokens, recent structures, UI Guidance, IBR artifacts, Mockup Gallery, and research based on product/workflow/data/platform fit.

Generate accessible color systems and solve contrast, instead of guessing hex values. Use when picking or fixing colors for any UI, chart, diagram, doc, or artifact — "what colors should this use", "is this readable", "this text fails contrast", "make a dark theme", "pick an accent". Also use before shipping any palette, to PROVE contrast rather than assert it. NOT for choosing layout, typography, or component structure.

Use when a Build Loop run or Git worktree touches mutable non-Git state: SQLite or file-backed databases, PostgreSQL databases/schemas and migrations, generated search/vector indexes, Docker volumes or service projects, mutable file trees, caches that cannot be rebuilt safely, or external cloud/account namespaces. Inventories each surface, selects per-worktree, shared-readonly, shared-serialized, or external-namespaced isolation, registers it in the run data manifest, validates collisions before writes, and requires a terminal disposition at closeout.

Use when choosing a model tier or segment for a subagent, deciding a role descriptor (segment + tier) in frontmatter, or escalating mid-flow. Covers the two-axis taxonomy (work-role segment × 7-rung capability ladder) — Opus/Sonnet/Haiku are Anthropic-default mappings; selection is provider-portable and data-driven. NOT for benchmarking models head-to-head on a task — use `model-bakeoff`.

Orchestrated build loop for multi-step code work. TRIGGER on verb language ('build', 'implement', 'create', 'add', 'ship', 'wire up', 'integrate', 'refactor', 'migrate', 'rewrite', 'replace') OR symptom language ('fix', 'broken', 'doesn't work', 'isn't loading', 'not displaying', 'missing', 'should show', 'needs to', 'make it', 'show this differently') OR any task touching 2+ files, adding/removing an endpoint, crossing an architectural boundary, or attached screenshots of a bug. SKIP one-line edits, pure Q&A, conversational clarifications, status checks, and trivial typos/renames.

Use when the build needs to automate a macOS .app without touching the hardware cursor, or the user asks to "click through the app" or "test the UI headlessly". Drives running apps via Accessibility API; self-contained Swift binary — no IBR, Playwright, or Appium required.

Audit and evolve repository structure safely across the full maintenance lifecycle: repository topology and scope, application and build-system profiles, module and folder boundaries, canonical source-of-truth and monorepo decisions, sibling consolidation, generated/build artifact retention, worktree and branch hygiene, local-versus-remote state, verified closeout into local main, and preparing a repository for open-source or external distribution (personal-content and secret scrub, internal-artifact de-tracking, .git bloat and history-leak review, distribution signing/notarization). Use when asked how any kind of repo should be structured or maintained, what application/build profile it has, why it has duplicate source or many build directories, whether repositories or modules should split or merge, to clean generated artifacts or a polluted data store (a log, telemetry, cache, or dataset corrupted by test/non-production writes), to review open commits/worktrees/branches/stashes, to merge and close completed

Operational lessons for maintaining a Claude Code plugin marketplace. Load whenever making changes to `.claude-plugin/marketplace.json`, bumping a plugin version that's listed in a marketplace, adding/removing plugins from a marketplace, or troubleshooting "add marketplace" failures. Prevents schema rejection, stale manifests, CDN-lag false negatives, and install-command drift.

How to publish a public npm package to GitHub Packages for any project under the @tyroneross scope — Claude Code plugins, build-loop subsystems, prompt-builder, etc. Load whenever the user asks to "publish to GitHub Packages", "publish the plugin", "ship as an npm package", "make this installable", "release this to npm", or whenever bumping a package version that needs to ship. Covers single-package repos (build-loop pattern) and monorepo per-plugin publishing (RossLabs-AI-Toolkit pattern). Includes the GitHub Actions workflow template, the release ritual, and the verification checklist that catches the 403 / version-drift / scope-missing failures already seen in this account.

Use when building an evaluation harness for an LLM agent, RAG pipeline, or prompt — when the question is "is this actually working?" Encodes Anthropic and OpenAI's published 2025–2026 eval guidance — start with 20–50 real-failure tasks (not synthetic happy-paths), use three grader types in the right places (deterministic code / LLM judge / human review), measure pass@k vs pass^k for production reliability, grade outcomes not reasoning paths, calibrate the LLM judge against human review before trusting it, and detect eval saturation. Triggers on "build an eval", "build evals", "eval harness", "is this prompt working", "regression test for LLM", "how do I know my agent is better", "LLM judge", "pass@k", "compare two prompts on real data", "the prompt looks fine but production is broken". Vendor-neutral with starter templates.

Use when designing the surface of a tool that an LLM agent will call — function declarations, MCP server methods, OpenAI function-calling schemas, Anthropic tool_use definitions, or Agent SDK tools. Encodes Anthropic and OpenAI's published 2025–2026 guidance — namespaced names (service_resource_verb), semantically specific parameters (user_id over user), signal-dense return formats with a response_format knob, prompt-engineered error strings that guide the model toward the fix, and the anti-pattern of wrapping every API endpoint 1:1. Triggers on "design a tool for an agent", "write a function declaration", "MCP server tool", "function calling schema", "tool_use definition", "Agents SDK tool", "the agent keeps calling my tool wrong", "the agent loops on a tool error", "list-all is blowing up context", "tool descriptions". Cross-vendor; vendor wire-format notes in references/.

Use when building or operating a coding agent that has to work across multiple context windows or multiple sessions — when one shot won't fit. Encodes Anthropic's published 2025 harness pattern — initializer/coding split, claude-progress.txt + feature-list JSON + git commits as cross-context state, mandatory session-init protocol (read logs → review features → run e2e → start dev server), edit-status-only contract on the feature file. Triggers on "long-running agent", "multi-session agent", "context window keeps filling up", "agent forgets between sessions", "session handoff", "Codex to Claude Code handoff", "Claude Code to Codex handoff", "harness for an agent", "what does the agent read on session start". Bridges bookmark (session snapshots) and build-loop (single-session orchestration).

Use when designing a prompt or agent that will run at scale and you want to maximize prompt-cache hit rate. Encodes the cross-vendor invariant — order content static-to-dynamic so the cacheable prefix is stable — plus Anthropic's explicit cache_control breakpoint mechanics (4-breakpoint budget, 20-block lookback, 5m vs 1h TTL) and OpenAI's automatic prefix hashing (≥1024 token threshold, 128-token increments, 5–10min idle TTL, up to 24h on gpt-5.5+). Triggers on "prompt caching", "cache hit rate", "reduce LLM cost", "ephemeral cache", "cache_control", "prompt is slow on every call", "we're paying for the same tokens twice", "TTL", "cache breakpoint". Vendor-neutral spine plus per-vendor branches in references/.

Use when prompting a reasoning model — OpenAI o-series (o3, o4-mini, o5), GPT-5 with reasoning_effort, or Anthropic Claude with extended thinking enabled. This is a counter-skill — the default prompt-builder advice (CoT, role-priming, "think step by step") actively hurts reasoning models. Encodes OpenAI and Anthropic's published 2025–2026 guidance — zero-shot first, do NOT add chain-of-thought, do NOT role-prime, developer messages replace system, reasoning_effort is a tuning knob not a quality lever, "Formatting re-enabled" recovers structured output, preserve thinking blocks across tool turns on Anthropic. Triggers on "prompt an o-series model", "prompt o3 / o4-mini / o5", "GPT-5 reasoning_effort", "extended thinking on Claude", "reasoning model", "the o-series is being weird", "the model ignores my CoT prompt", "thinking blocks", "should I use chain-of-thought here".

Use when writing a system prompt for an LLM response that must cite sources, label every number with origin, and stay in a consistent voice. Composes three reusable rule blocks (citation contract, number-labeling, two-register voice) onto a base prompt and gives you the contract tests to keep them from drifting. Triggers on "build a system prompt", "make it cite sources", "add citations", "ground the output", "label numbers", "truthfulness rule", "system prompt for RAG", "system prompt for an audit view", "stop the model from making up numbers", "the model is citing things that aren't in the retrieval list". Designed for RAG, agentic Q&A, recommendation surfaces, and audit/explain views. Not for creative writing, code generation, or pure structured-extraction tasks.

Use when /ibr:build detects .mockup-gallery/ or IBR needs to tie a build to gallery-selected mockups. Reads ratings/selections, enforces approved target roles, preserves scratch-first workflow, and writes implementation completion. Filesystem only.

Use when validating iOS, watchOS, or macOS UI with IBR.

Use when working on iOS, watchOS, or macOS apps — .swift files, SwiftUI views, simulators. IBR's native scan workflow for touch targets, a11y labels, watchOS constraints, Fix Guide.

Use when scanning, auditing, comparing, or testing UI with IBR.

Use when planning or implementing UI with IBR.

Use when planning, designing, implementing, or auditing UI/UX with IBR in Codex. Provides the latest compact IBR guidance for Calm Precision, web archetypes, interaction states, mobile, content states, data visualization, Mockup Gallery targets, and imagegen approval gates.

Use when building cards, navs, forms, dashboards, modals, tables, or lists. Opinionated patterns with Calm Precision principles, spacing, accessibility, and anti-patterns.

Use when UI work includes charts, graphs, dashboards, KPIs, analytical search responses, metrics, tables, trend lines, rankings, distributions, or data storytelling. Provides chart-worthiness gates, chart routing, hierarchy, attribution, accessibility, and validation rules.

Use when user asks to show or generate an architecture diagram, visualize dependencies, export architecture docs, create architecture documentation, save architecture output, or create a mermaid diagram of their project.

Use when user asks to scan architecture, check dependencies, find outdated packages, show project structure, refresh architecture, or run a health check. Navgator scan, status, and staleness check.

Use when user asks to review architecture, check connections, navgator review, is this safe to merge, what did I break, or when scan detects architectural drift. Architectural integrity review of system flow and connections.

Use when user asks what breaks if I change X, impact of changing, what depends on, trace data flow, show connections, dependency graph, upstream/downstream, or safe to modify before refactoring.

Detect and analyze infrastructure components — Prisma models, env vars, queues, crons, deploy configs, field usage, and type validation

Use when user asks to install navgator, update navgator, set up navgator, launch the dashboard, or run navgator ui maintenance.

Use when user asks to audit, critique, review, or diagnose existing writing against the pyramid principle, Minto, MECE, or SCQA. Produces rule-by-rule findings with citations and ranked fixes.

Use when user asks to outline, structure, or restructure a long report, brief, proposal, whitepaper, strategy doc, or research writeup using the pyramid principle. Full nested key-line outline.

Use when user asks to structure, storyline, or outline a presentation, deck, slides, ghost deck, or talk using the pyramid principle. Produces slide-by-slide assertion headlines and key lines.

Use when user asks about pyramid principle, Minto, SCQA, MECE, governing thought, key line, vertical/horizontal logic, or answer-first structure. Canonical rule library for structured communication.

Use when user asks to draft an email, memo, executive summary, one-pager, BLUF, Slack update, or status note using the pyramid principle. Answer-first prose under ~500 words with SCQA and key line.

Use when planning or producing product marketing content/video for any software product — "plan a marketing video", "app store preview", "product hunt launch", "explainer vs demo", "LinkedIn/YouTube video for <product>", "launch video", "what video should I make for <app>", "ongoing content plan", "changelog/release video", "readme demo GIF", "repurpose this into shorts", or when Spectra is about to capture/produce marketing assets and should tailor them to the product's audience, channel, and funnel stage. Turns product × audience × goal into a concrete plan (video type, length, channel, story-spine script, specs, polish style) and routes production to Spectra's capture/record/polish tools.

Use when a video's craft needs tuning — "this video looks/feels off", "how should this video be paced", "what color/typography/font size for video", "music or voiceover pace", "cut rate", "why isn't my video engaging", "make it feel premium" or "make it feel authentic", "kinetic text", "safe zones", or when Spectra is polishing/rendering a clip and needs concrete design settings (color, type, sound, pacing). Turns the emotional/visual/sound/pacing science into font-size floors, cut-rate targets, color rules, and Spectra polish parameters.

Use when the user asks to produce a polished demo video from a screen recording, highlight key moments in a recording, add spotlight focus to video segments, create an agent demo video, or trim and caption a screen capture for sharing.

Use when the user asks to walk me through a UI, demo this flow, show me how a feature works, capture a walkthrough, or navigate a web/macOS app with Spectra.

Use when the user asks to capture screenshots, record a demo or video, create marketing assets, or document a feature visually from a running application.

Use when the user asks to "debug this", "fix this bug", "investigate error", "diagnose", "root cause", or reports a crash/exception/failure. Memory-first workflow checks past incidents before investigating.

Use for deep iterative debugging when memory lookup returns LIKELY_MATCH/WEAK_SIGNAL/NO_MATCH, a fix didn't hold, or the user asks for root cause analysis. Not for known fixes or trivial issues.

Use when the user asks to "add logging", "add tracing", "improve observability", "OpenTelemetry", "structured logging", or reports silent failures or no runtime visibility. Generates stack-appropriate logging with optional OTel.

Use when working in a repository that uses Rally/Agent Rally Point for cross-agent coordination, especially at session start, before editing files, when deciding what to do next, handing work to another agent, recording facts/artifacts/decisions, resolving blockers, or coordinating with other coding agents through the `rally` CLI.

Ultra-light per-task quality loop for Rally Flow / dynamic-workflows fan-out. Wrap each fanned-out task in assess → plan → execute → mini-judge to raise accuracy, with zero tools or scripts. Use when running a workstream task and you want it to actually meet its validation + output contract, not just "be done".

Use when fanning out work across multiple agents, running a dynamic workflow, coordinating parallel subagents, or splitting a workstream across hosts, terminals, or machines through Agent Rally Point. Defines the workstream descriptor + task-packet protocol and the per-task rally coordination loop. Host-neutral — works for any coding agent.

Use to create scratch-first HTML mockups, optionally use image generation for visual assets, launch/review mockups, check selections, align mockups to routes/screens, or implement approved designs from the mockup gallery.

Use to create HTML mockups for review, check design selections/ratings, or implement approved designs. Triggers on "mockup", "design review", "gallery", "rate mockup", "accepted designs".

Use to create HTML mockups for review, check design selections/ratings, or implement approved designs. Triggers on "mockup", "design review", "gallery", "rate mockup", "accepted designs".

Use BEFORE writing API config, env setup, auth, LLM client init, or debugging an external service. Returns authoritative source URLs, latest version, and deprecation warnings. Triggers on any named library/API/tool.

Contract for external plugins (build-loop, debugger, research) to consult api-registry during API setup or debug. Reads ~/.api-registry/registry.db; exits silently if absent.

Use when the user asks how a library/API/tool works, requests docs for a named service, or asks about current models/versions/endpoints. Routes via api-registry to Context7 or WebFetch for authoritative answers.

Use when the user asks to "make this an iOS app", "convert to native", "port to Swift", "SwiftUI version", or migrate a Replit app to native iOS/macOS.

Use when the user asks to "migrate to web", "deploy to Vercel", "move to Cloudflare", "get off Replit", or convert a Replit app to standalone/self-hosted production.

Use to "scan this Replit app", "analyze for migration", "check Replit dependencies", or assess migration readiness. Auto-triggers when a .replit file is present.

Design, evaluate, or rebuild agentic harnesses — tool use, permissions, workflow state, memory, evals, observability, multi-agent, framework selection. Activates on harness-gap symptoms too.