llm_skills

Automated project implementation orchestrator that drives feature-driven development from a single initial prompt through to completed code. Use this skill when the user invokes /init-project, /init-features, /add-feature, /continue-tasks, /continue-new-session, /iterate-tasks, /review-tasks, /update-tasks, /analyze-features, or /reinit. Also trigger proactively when docs/INITIAL_PROMPT.md exists and the user says anything like "move forward", "keep building", "what's next", "continue the implementation", or "start working on the project", AND when the user says "set up project management", "bootstrap a new project", "initialize the project workflow", or "make sure agents follow the project plan". This skill manages the full lifecycle: scaffolding a new project with enforcement artifacts, extracting feature specs via interview, generating phased implementation plans, spawning typed agents to execute tasks, monitoring completion sentinels, recovering from failures, and archiving finished work — all driven by m

Generate a copy-ready prompt that a fresh session (Claude Code, Codex, or Gemini) can paste in to immediately begin work on the next recommended action that was identified in this session's most recent recap. Reference-style — names file paths with short excerpts, no large inline content.

Self-perpetuating, subagent-driven iteration of the project pipeline. One invocation: (1) merges any pending PR left by the previous turn, (2) dispatches the recap-recommended next action as a fresh subagent so each unit of work gets a clean context, then (3) emits a copy-ready prompt for the action that comes AFTER the one just completed. Designed to pair with /loop for unattended runs and to minimize parent-context growth.

Run the full project orchestration loop with approved-spec gating, dependency ordering, verified completion, and deterministic failure/blocked handling

Use when starting a new project from scratch or retrofitting the project-manager workflow onto an existing repo. Scaffolds the docs/ tree, copies canonical templates, installs agent-enforcement artifacts (AGENTS.md, CLAUDE.md fragment, pre-commit guard, Claude Code hook, PR template, ROADMAP), and seeds INITIAL_PROMPT.md. Idempotent: re-running only fills gaps.

Dry-run project status report. Produces a read-only snapshot of feature specs, dependency blockers, plans, active tasks, stale work, verification backlog, and open issues without spawning agents or modifying files.

Reconcile active task files with plans without spawning new workers, including success, failure, blocked, malformed, and stale states

Domain expertise for harvesting comments and threads from a YouTube video using yt-dlp. Sub-skill of `youtube-extraction`. Use when the parent skill needs comments as a primary source for filenames, repo URLs, corrections, and creator follow-ups that don't appear in the video itself.

Domain expertise for generating the final markdown report that summarises a YouTube extraction operation — files reconstructed, evidence trail, gaps, and follow-ups. Sub-skill of `youtube-extraction`.

Domain expertise for stitching multi-frame OCR evidence into canonical source files on disk — without duplicating overlapping lines — and for placing each file at a sensible path inside the calling repo. Sub-skill of `youtube-extraction`.

Domain expertise for visually inspecting extracted video frames to identify which ones show file content (IDE panes, terminal cat, slides, READMEs) and capture path, language, and visible line ranges. Sub-skill of `youtube-extraction`.

Domain expertise for slicing a downloaded YouTube video into image frames using ffmpeg — with sampling strategies tuned to the downstream task (file reconstruction, PRD evidence, diagram capture). Sub-skill of `youtube-extraction`.

Marko is a grumpy senior code reviewer who critiques code with zero praise inflation and never suggests fixes. Invoke Marko whenever the user addresses him by name to ask for a code opinion — "marko?", "what do you think marko?", "hey marko", "marko review this", "marko take a look", "marko thoughts?", or any message where "marko" is used as a vocative asking for feedback on code. Marko reviews the current diff or recent edits, issues a verdict (catastrophe / bad / mediocre / ok), and lists concrete file:line gripes another agent can act on. Do NOT invoke when "marko" appears as a person's name in passing ("marko from the design team"), only when he is being addressed. Always invoke when addressed, even if the user doesn't say "review" or "code" explicitly.

Full lifecycle management of LLM skills across the workstation — finding, archiving, installing, updating, and importing skills with their complete bundles (sub-skills + companion commands). Use when the user invokes /find-skills, /sync-skill, /install-skill, /update-skill, or /import-skill. All user interaction MUST go through the AskUserQuestion tool — never free-form text, never inline markdown questions.

Domain expertise for obtaining a timestamped transcript of a YouTube video — first via YouTube's own auto-subs/CC, falling back to local Whisper transcription of the downloaded video. Sub-skill of `youtube-extraction`. Use when the parent skill needs spoken-word evidence aligned to timestamps.

Use when the user is hitting Claude usage limits, burning through tokens fast, running a long conversation, or asks how to use Claude Code more efficiently. Triggers on phrases like "hit my limit", "running out of tokens", "usage limit", "save tokens", "reduce usage", "am I wasting tokens", "this chat is getting long", "which model should I use". Diagnoses the current session, runs a real token-usage report from local JSONL logs, and applies Dubi's 11 rules for reducing Claude usage.

Domain expertise for downloading a YouTube video locally at a resolution suitable for later frame OCR. Sub-skill of `youtube-extraction`. Use when the parent skill needs to acquire the source video file before any frame, transcript, or content analysis.

Decide what the agent should do next in the current repository. Use this skill whenever the user asks "what next?", "what should I work on?", "where did we leave off?", "what's on the backlog?", "help me pick the next task", or invokes /what-next or /what-next-update. Also trigger proactively when the user returns to a project after a break and wants direction. The skill inspects repo structure, detects the project-management framework in use, reads or builds a backlog, prioritises pending work using a weighted heuristic, and — after the user picks from the top three — hands the chosen task off to the right specialist agent. All user interaction goes through AskUserQuestion; results are cached to docs/what-next.md so subsequent runs skip re-analysis.

React composition patterns that scale. Use when refactoring components with boolean prop proliferation, building flexible component libraries, or designing reusable APIs. Triggers on tasks involving compound components, render props, context providers, or component architecture. Includes React 19 API changes.

React and Next.js performance optimization guidelines from Vercel Engineering. This skill should be used when writing, reviewing, or refactoring React/Next.js code to ensure optimal performance patterns. Triggers on tasks involving React components, Next.js pages, data fetching, bundle optimization, or performance improvements.

Full lifecycle management of LLM skills across the workstation — finding, archiving, installing, updating, and importing skills with their complete bundles (sub-skills + companion commands). Use when the user invokes /find-skills, /sync-skill, /install-skill, /update-skill, /import-skill, /push-skill, /search-skill, /audit-skills, or /backfill-diagrams.

Full lifecycle management of LLM skills across the workstation — finding, archiving, installing, updating, and importing skills with their complete bundles (sub-skills + companion commands). Use when the user invokes /find-skills, /sync-skill, /install-skill, /update-skill, /import-skill, /push-skill, /search-skill, /audit-skills, or /backfill-diagrams.

React Native and Expo patterns, platform-specific code, list performance, animations, and navigation best practices