Guides a founder through the full Y Combinator batch application end-to-end. A 10-phase workflow that captures the live YC form, profiles the founders, stress-tests the idea via an embedded grill loop, runs a mandatory 5-agent parallel external research pass on the startup, drafts every form field with anti-pattern and accepted-example checks, produces founder-video bullet notes (no script), runs a final adversarial gate, generates paste-ready submission answers, unlocks an interview-prep simulator after invite, and supports reapplicant delta tracking and post-decision post-mortems. Writes a documented markdown trail under a user-chosen workspace. Use when a founder wants to prepare a YC batch application, build their founder video, drill mock YC interview questions, or reapply with delta evidence. Don't use for pitch-deck design unrelated to YC, generic startup advice without applying, or post-funding work.
Creates and maintains Knowledge Base topics from YouTube channels by resolving recent or full uploads, scaffolding yt-channels topics, ingesting transcripts through kb ingest youtube, and validating/indexing the result. Use when turning a YouTube channel into a Karpathy KB topic. Do not use for single-video ingestion, general video summaries, or non-YouTube sources.
Authors engineering blog posts end-to-end: launch deep-dives, incident postmortems, architecture migrations, performance case studies, tutorials, AI/agent system writeups, security disclosures, and research-to-product translations. Picks the correct archetype, plans the abstraction ladder, enforces an evidence cadence (diagrams, benchmarks, profiles, traces, code, ablations), tunes voice against publisher house styles (Datadog, Vercel, GitHub, AWS, Meta, Cloudflare, Jane Street), and runs a pre-publish gate for narrative momentum and disclosure ethics. Use when drafting a new engineering post, restructuring a draft that feels flat, deciding which evidence form belongs where, validating that depth and product context are balanced, or preparing a postmortem, migration, or performance narrative for external publication. Do not use for API reference documentation, README authoring, marketing copy, release notes, generic SEO content, ghost-written executive thought leadership, or non-engineering long-form essays.
Dispatches scoped-write explorer agents in parallel for general research and exploration of any codebase, topic, or domain. The operator passes --path (output directory), --agents (parallel count), --prompt (research question), and optionally --ide, --model, --reasoning to control the Compozy runtime. The parent scouts the territory first to divide work, then invokes `compozy exec --agent explorer` N times in parallel (via the harness's async/background facility); each invocation writes one analysis file at <path>/analysis/NN_analysis_<slug>.md following a seven-section schema. The parent then synthesizes <path>/analysis/summary.md. The explorer agent lives in the Compozy global registry at ~/.compozy/agents/explorer/AGENT.md and is installed by the bundled script when absent. Use when running parallel multi-area research that must produce written artifacts. Do not use for competitor-only research already covered by cy-research-competitors, single-file lookups answerable by Explore, or edits to existing code.
Audits AI-implemented work for honest completion. Runs independent-evaluator checks against task artifacts, transcripts, tests, CI evidence, requirement-to-test mapping, status front matter, and quality gates; flags skipped tests, weakened assertions, mock-only confidence, snapshot drift, happy-path-only coverage, flaky retries, and status/evidence mismatches. Use when validating completed Compozy tasks, AI-authored PRs, or codex-loop iterations. Do not use for real-user QA, persona/journey testing, exploratory charters, or product usability sessions; use qa-execution for those.
Guides creation and modification of domain feature systems organized under a systems/ directory. Covers directory layout, API service layer patterns, TanStack Query hooks (queries, mutations, optimistic updates), React context and XState store conventions, hook organization, and public API barrel exports. Use when adding a new domain system, extending an existing one, or fixing bugs in a system-layer codebase. Don't use for generic React component work, backend API implementation, or codebases not organized around a systems/ domain pattern.
Deep architectural audit focused on finding dead code, duplicated functionality, architectural anti-patterns, type confusion, and code smells. Use when user asks for architectural analysis, find dead code, identify duplication, or assess codebase health. Don't use for style/formatting issues, performance profiling, security audits, or feature-level code review.
Build terminal user interfaces with Go and Bubbletea framework. Use when creating TUI apps with the Elm architecture, dual-pane layouts, accordion modes, mouse/keyboard handling, Lipgloss styling, and reusable components. Includes production-ready templates, effects library, and battle-tested layout patterns from real projects. Don't use for plain-text CLI scripts without an interactive UI, web/desktop GUIs, or non-Go terminal frameworks (Ink, Textual, Ratatui).