↓ Skip to main content

Agent Skills

Curated instructions and workflows to level up your AI coding agents.
🔍
🪦 Graveyard

swarm-coding

Hierarchical multi-agent engineering workflow that divides multi-system codebases across a root Coordinator, domain Tech Leads, and task-scoped Specialists. Uses spec-first contracts, Single Writer file ownership, and a Map-Reduce integration pass to prevent context bloat and edit collisions. Use for large full-stack builds or cross-domain refactors spanning independent subsystems such as backend, frontend, data, and QA.

double-diamond

Universal multi-agent orchestration workflow based on the Double Diamond framework (Inception -> Discovery -> Definition -> Development -> Delivery). Coordinates parallel subagents with context isolation to separate problem-space research from solution-space implementation. Activate for complex, high-ambiguity initiatives across software engineering, in-depth research, long-form writing, legal analysis, or product strategy requiring structured human alignment gates.

game-design

Interactive 2D game design workflow and Game Design Document (GDD) authoring guide. Structures co-creation through one-decision-at-a-time interviews covering genre, camera projection, core loops, win/loss rules, controls, 2D HD or pixel-art sprite pipelines, and procedural or AI audio and cutscenes into GDD.md. Activate when conceptualizing a 2D game, defining mechanics, or authoring a Game Design Document.

parallel

General-purpose task parallelization and agent scaling engine. Orchestrates concurrent subagents for research, analysis, writing, coding, planning, and evaluation across numeric budgets (DOP) or automatic sizing (auto). Enforces non-blocking coordination, disjoint task partitioning, and agent racing or tournament selection. Activate when the user requests parallel execution, specifies a concurrency budget, or asks for concurrent subagents.

uno-reverse

Contrarian simplification, red-teaming, and Occam's Razor review for software architecture, PRDs, agent workflows, and technical designs. Challenges feature creep, speculative abstractions, and bloated specs by proposing minimum viable primitives that deliver 90% of value with 10% of moving parts. Delegates reviews to an isolated subagent by default to avoid confirmation bias. Activate when pruning bloated designs, red-teaming specs, or finding the simplest path to production.

godoctor

Developer tooling and architectural safety rules for Go. Orchestrates high-level Go toolchain and static analysis utilities (gofmt, goimports, go vet, modernize, golangci-lint, deadcode), guards edits with compiler rollback gates, eliminates blind spots via Selene mutation testing, and enables fast test and coverage analytics with TestQuery SQLite queries. Activate when writing or refactoring Go code, fixing compilation or test failures, auditing test thoroughness, or enforcing idiomatic Go standards.

nano-banana

Conversational image generation and multimodal editing with Google Nano Banana models via uv run scripts/banana.py, the gemini-api CLI, and Google Gen AI SDKs. Generates up to 4K visuals, panoramas, character sheets, and chroma-key sprite grids with multi-image reference consistency and Search grounding. Activate when generating or editing images, sprites, or art, or when asked to "banana" an image.

deslopify

Editorial guide and rewriting workflow for removing AI cliches, tropes, and formulaic filler from prose. Flags overused vocabulary (such as delve, tapestry, seamlessly), negative parallelism, dramatic countdowns, reasoning throat-clearing, and fractal summaries while preserving the author's voice. Activate when auditing prose for AI tells or polishing drafts to read naturally.

ebitengineer

Engineering architecture and best practices for building 2D games in Go using Ebitengine (v2). Covers modular game architecture, scene state machines, virtual resolution scaling, delta-time physics, zero-allocation draw loops, sprite sheet and AnimationController playback, 4-byte aligned 44.1 kHz stereo PCM audio, Kage shaders, and WebAssembly builds. Activate when building, optimizing, or debugging 2D games in Go, or compiling Ebitengine games to WebAssembly.

seo-optimizer

Technical SEO and Generative Engine Optimization (GEO) guide for developer docs and engineering blogs. Splits human summary hooks from search description metadata, validates TechArticle JSON-LD, configures AI crawler robots.txt rules, and generates llms.txt indexes. Activate when auditing technical SEO, tuning pages for AI search grounding, or structuring frontmatter metadata.

social-copy

Editorial workflow and platform playbooks for developer social media copy and campaigns. Extracts technical evidence from git history or author interviews, builds a canonical narrative (CANONICAL.md), adapts posts sequentially for LinkedIn, X, Bluesky, Instagram, Reddit, and Threads, and checks for AI cliches. Activate when drafting or auditing developer social posts and launches.

vibe-game-developer

Workflow router and orchestrator for building 2D games in Go with Ebitengine. Coordinates game design and GDD creation (game-design), engine architecture and WebAssembly builds (ebitengineer), 2D HD and pixel-art sprites (sprite-design and nano-banana), procedural and AI audio and video cutscenes (procedural-sfx, procedural-music, lyria, and gemini-omni), Go verification (godoctor), and multi-agent scaling (swarm-coding, parallel).

inverted-pyramid

Structural guide for organizing technical articles, developer documentation, and README files with the inverted pyramid model. Places summaries and runnable quickstarts first, followed by usage workflows, architecture, and compliance. Activate when structuring technical posts, writing or refactoring README files, or designing scannable documentation hierarchies.

google-oss

Compliance and licensing guide for Google Open Source repositories and personal open-source projects published by Googlers. Enforces source file license headers (such as Apache-2.0) using addlicense, verifies license files, inserts required non-official product disclaimers, and audits pre-release compliance. Activate when preparing Google Open Source or Googler personal projects for release, applying copyright headers, or adding required disclaimers.

buffer

Manage, draft, schedule, and publish social media content across connected channels (LinkedIn, X/Twitter, Bluesky, Threads, and Instagram) using the Buffer CLI (@bufferapp/cli). Covers channel discovery, dry-run validation, approval-gated queue scheduling, draft ideas, and live schema inspection. Activate when scheduling social media posts, inspecting Buffer channels, or automating social publishing workflows.

lyria

Generate 44.1 kHz stereo music, full songs with lyrics, and instrumental tracks from text prompts or reference images using Google Lyria models (lyria-3.5, lyria-3-clip-preview, lyria-3-pro-preview) via uv run scripts/lyria.py, the gemini-api CLI, or Google Gen AI SDKs across Gemini Enterprise / Vertex AI (global) and Gemini API. Activate when composing soundtracks, generating songs, or creating audio from images.

buffer-analytics

Collect and analyze social media data from Buffer in a local SQLite database. Stores your full post history and metrics across connected channels (such as LinkedIn, X/Twitter, Bluesky, and others) so you can run SQL queries or view reports on engagement, clicks, and views. Activate when you need to analyze social media performance, find the best days or times to post, identify top-performing content, or query Buffer data with SQL.

google-analytics

Collect and analyze Google Analytics 4 (GA4) website data in a local SQLite database. Stores pageviews, active users, reading dwell time, traffic sources, and outbound clicks so you can run SQL queries or view reports on site performance. Activate when analyzing website traffic, measuring reader engagement and dwell time, evaluating the impact of site updates or milestones, or querying Google Analytics with SQL.

skill-analytics

Mine and analyze Agent Skill usage, prompt-tax ROI, and trigger attribution from conversation transcripts in a local SQLite database. Scans conversation transcripts and installed skill tiers (built-in, plugin, global, workspace, and catalog) to detect zero-use shelfware and tune activation descriptions. Activate when analyzing skill usage over time, auditing system-prompt token overhead, measuring autonomous vs. explicit (::skill or //skill) triggers, or querying conversation skill events with SQL.

search-analytics

Collect and analyze Google Search Console organic search data in a local SQLite database. Stores clicks, impressions, click-through rates (CTR), average ranking positions, and landing pages so you can run SQL queries or view performance reports. Activate when analyzing Google Search traffic, tracking keyword rankings, finding SEO content optimization opportunities, or querying Search Console data with SQL.

intercom

Cross-project and cross-session collaboration between Antigravity agents using the agentapi CLI and conversation_summaries.db. Finds active peer sessions across workspaces, looks up target project IDs, and routes messages to live sessions in other projects with return-address headers. Activate when collaborating between agents in different projects, handing off tasks across workspaces, finding peer session IDs, or sending cross-project messages with agentapi.

abstract

Philosophical and inductive knowledge-distillation workflow that extracts general principles, causal laws, and portable concepts from concrete facts, evidence, and examples. Applies Aristotle's essence-versus-accident split, Bacon's inductive ascent, Peirce's concept naming, and Popper's flip test to turn raw facts into high-value knowledge without hollow platitudes. Activate when abstracting ideas, distilling lessons, or synthesizing principles from evidence.

brainstorm

Divergent ideation and option-generation engine that breaks LLM mode collapse using symmetry-breaking algorithms and optional parallel subagent dispatch. Applies taboo deflection, assumption inversion, tension breaking, cross-domain transfer, and testable bets to generate 3 high-conviction proposals per run, or fans out identical instructions across N subagents before reducing consensus and outlier breakthroughs. Activate on brainstorm(N) or when exploring non-obvious solutions, architectures, or creative directions.

design-docs

Author and manage Request for Comments (RFC) proposals and Architecture Decision Records (ADRs). Covers collaborative design proposals, trade-off analysis, spike evidence, proposal lifecycle states, and immutable decision logs across design/rfc/ and design/adr/. Activate when proposing features, evaluating architectural options, or recording engineering decisions.

discovery

Scientific and procedural problem-space exploration workflow that widens the investigative aperture across any domain to surface blind spots, hidden constraints, overlooked outliers, and unknown unknowns before narrowing on a choice. Frames and quantifies findings using a 6-tier evidence hierarchy. Activate when exploring a problem space, investigating systems or markets, hunting for hidden gems, or uncovering what an initial framing missed.

gemini-omni

Generate, animate, edit, interpolate, extend, and upscale video clips (3 to 15 seconds, 360p to 4K at 24 FPS) with native synchronized audio using Gemini Omni 1.2 Flash (gemini-omni-1.2-flash) via uv run scripts/omni.py and the Google Gen AI Interactions API across Gemini Enterprise / Vertex AI (global) and Gemini API. Activate when creating videos from text or images, editing or extending clips, or upscaling video resolution.

less-is-more

Craft philosophy focused on removing clutter, cutting visual pollution, and doing one core thing at a high standard instead of shipping ten mediocre extras. Guides product scope, interface design, software architecture, and writing toward clarity, restraint, and excellence. Activate when simplifying cluttered designs, trimming feature creep, cleaning up noisy interfaces, or sharpening focus on what matters most.

pagespeed-insights

Run and extract Google PageSpeed Insights Mobile and Desktop audits via the v5 REST API (runPagespeed) or pagespeed.web.dev SSR window variables. Returns CrUX Core Web Vitals (LCP, INP, CLS) and Lighthouse lab diagnostics (TBT, FCP, Speed Index) without DOM snapshots. Activate when auditing a live URL or pagespeed.web.dev report or fixing Core Web Vitals.

procedural-sfx

Pure-code procedural game sound effect (SFX) synthesis engine and .sfx DSL for Ebitengine. Compiles 4-byte aligned 16-bit 44.1 kHz stereo PCM cues in memory with PolyBLEP anti-aliasing, 18 archetype presets, duty and SVF filter sweeps, Galois LFSR chiptune noise, 2-operator FM, soft-clipping, boundary micro-fades, and acoustic QA validation.

show-me-the-money

Mine project conversation history and git logs using a Map-Reduce workflow with parallel Gemini 4 Hydrogen agy CLI workers to recover unimplemented features. Scans all sessions in the current project by default, running one worker per session in the Map step and resolving the latest snapshot in the Reduce step. Activate when looking for forgotten features, mining past sessions for unfinished work, or asking "what did I forget to build?".

spike

Pragmatic Agile spike workflow for retiring technical, design, or functional risk before committing to implementation or estimates. Applies Extreme Programming risk-reduction rules with a right-sized toolkit, spanning spec and source audits, blast-radius traces, state-table walk-throughs, and throwaway probes, to deliver a fast, proof-backed Go/No-Go/Pivot verdict. Activate when running a spike or resolving a blocking unknown.

sprite-design

Guide and .sprite engine for 2D game sprites, tilesets, and frame animations in Ebitengine. Covers the two-step AI character sheet to chroma-key grid pipeline with nano-banana, normalization into .sprite, .aseprite, .png, and .json containers, procedural SDF sprites, visual QA checks, and zero-allocation AnimationController state machines.