jwynia/agent-skills
18 skills17.3k installs1.7k starsGitHub
Install
npx skills add https://github.com/jwynia/agent-skillsSkills in this repo
1Frontend DesignFrontend Design is an agent skill packaged as a component specification schema for template generation. Solo builders use it when they want consistent, agent-driven UI scaffolding instead of one-off chat snippets that ignore framework and styling constraints. The spec forces PascalCase component names, picks a component archetype, selects a target framework among React, Vue, Svelte, or plain HTML, and pairs a styling approach such as Tailwind or CSS modules with an optional aesthetic profile including animation intensity and dark mode. That structure helps Claude Code, Cursor, and Codex emit components that align with a deliberate visual direction—minimal, brutalist, glassmorphism, and others—rather than generic defaults. It fits the Build phase when you are composing dashboards, marketing surfaces, or product chrome and need repeatable prompts that encode design decisions up front.2.3kinstalls2Requirements AnalysisRequirements Analysis is an agent skill that helps solo builders move from vague intent to grounded, validated needs. It diagnoses requirements-level failure modes—such as jumping straight to features, missing a problem statement, or copying competitors without understanding why—and applies structured questions and interventions rather than filling a template. The skill frames requirements as hypotheses about what will solve a problem, aligning discovery with whether those hypotheses match reality. It supports Jobs-to-be-Done style interviews, naming concrete users, and tracing ideas to originating frustrations. Use it whenever you are tempted to build before you can say who suffers without the product and what they do today instead. It pairs naturally with later scoping and planning skills once the problem is legible.2.1kinstalls3Godot Best Practicesgodot-best-practices documents how solo and indie game devs should structure global singleton managers in Godot 4.x using autoloads registered in Project Settings. The skill emphasizes thin service-oriented autoloads—audio, persistence, scene flow—not monolithic game rules, and supplies copy-paste GDScript templates with signals, private state, lifecycle setup, and a documented public API. The included AudioManager walkthrough shows buses, pooled SFX players, and music change signals, which is the kind of repeatable pattern agents often get wrong in one-off chats. On Prism it sits squarely in the Build phase for Godot titles, giving builders a procedural template rather than an external API integration, so you can standardize manager code across a project before testing and ship.1.8kinstalls4Presentation DesignPresentation Design is a diagnostic and assistive agent skill for building talks and slides that prioritize audience understanding over presenter convenience. Solo builders use it when structuring a launch webinar, tightening a pitch, or getting feedback on decks built in reveal.js, PowerPoint, Keynote, or Google Slides. The skill maps common failure modes—walls of text, bullet sprawl, cognitive overload—to concrete fixes like one concept per slide and assertion-evidence layout. Phase 1 planning forces clarity on audience, a single main message, and what can be cut. It fits Launch distribution work first, but the same frameworks help Validate pitching and Grow content education. Invoke it before you lock slide order or after a first draft when you need structured critique rather than generic “make it prettier” advice.1.7kinstalls5Web Search TavilyWeb Search Tavily is an agent-adjacent CLI skill that wraps Tavily’s AI-optimized search API in a small Deno script you can invoke from Claude Code, Cursor, or any shell workflow. Solo and indie builders use it when they need reliable web evidence—not generic chat hallucinations—for competitor scans, audience research, pricing benchmarks, or technical fact checks during early journey phases. The script supports configurable max results, search depth, topics, and optional answer synthesis so agents can consume concise passages designed for downstream reasoning. It is not a full browser automation stack; it is a focused network integration that assumes you supply an API key and accept outbound calls to Tavily. Pair it with note-taking or synthesis skills after you have raw hits. Security and quota behavior depend on your Tavily account; review Prism’s Security Audits panel on this listing before piping queries from CI or shared machines.1.1kinstalls6Pdf Generatorpdf-generator is a Deno-based agent skill for creating and manipulating PDF files in code. Solo and indie builders use it when invoices, reports, certificates, or filled government forms must ship as PDFs instead of HTML or spreadsheets. It supports two paths: template mode to fill and overlay existing PDFs, and scratch mode to assemble documents from JSON specs. Typical jobs include merging annexes, stamping watermarks, and pulling text or field metadata for validation pipelines. The skill expects Deno on the machine and local read/write access to inputs and outputs. It is a poor fit when someone only wants to view a PDF, needs OCR on scans, or requires rich typographic layout—those cases need viewers, OCR tools, or HTML-to-PDF stacks. For agent workflows, it turns structured agent output into downloadable, shareable PDF artifacts without a separate document SaaS.865installs7Godot Asset GeneratorGodot Asset Generator is a generator skill for indie game builders using Godot 4.x who want AI-assisted 2D art without leaving a documented pipeline. It walks concept and style consistency through API calls (OpenAI, Replicate, or fal.ai), sheet preparation, and Godot import configuration so sprites and tilesets land correctly in the filesystem and editor. Solo builders hit it when manual pixel work is too slow but they still need cohesive sets—platformer characters, top-down tiles, menu chrome. You need Deno plus the provider key matching your chosen backend. Skip it for pure hand-drawn art workflows or any 3D mesh/texture work; the skill explicitly excludes those paths. Intermediate comfort with Godot’s import defaults and generative prompt craft improves outcomes.812installs8Story CoachStory Coach is an assistive writing skill for fiction authors who want an agent to behave like a human coach, not a ghostwriter. Its license and frontmatter mark it as diagnostic and assistive: the agent may ask probing questions, explain story frameworks, diagnose why a beat feels flat, and offer approaches—but it must refuse to produce narrative prose, dialogue, scene text, unsolicited outlines, or lore dumps. Solo builders shipping creative products (serial fiction, game writing, brand stories) use it whenever they are stuck on character motivation, structure, or revision without outsourcing their voice. The skill fits any phase where you are making creative commitments before locking copy: brainstorming angles in Idea, scoping a novella in Validate, drafting in Build-adjacent writing sessions, or sharpening launch copy that must stay in-character. Invoke it when you have something on the page or a clear intent and need guided reflection rather than generated text.757installs9Web SearchWeb Search is a lightweight agent skill that teaches your coding agent how and when to call its built-in WebSearch capability instead of guessing from stale training data. It is aimed at solo and indie builders who ship with Claude Code, Cursor, Codex, or similar environments where web lookup is already available. Use it whenever a task depends on recent events, version-specific docs, pricing pages, competitor moves, or verifiable citations—you get a repeatable pattern for query formulation and scope control without provisioning another search vendor. The skill deliberately stays out of the way for pure repo work and points you to a Tavily-based alternative when you need richer filtering. For Prism’s journey model, it behaves as journey-wide research glue: the same discipline applies in Idea (market scan), Validate (fact-check claims), Build (framework APIs), Ship (security advisories), Launch (SEO snippets), Grow (trend monitoring), and Operate (incident context)—always as an adjunct to judgment, not a replacement for reading primary sources.725installs10Docx Generatordocx-generator (as documented via analyze-template.ts) is an agent skill path for solo builders who automate Word documents from repo-local tooling. It reads a .docx as a zip bundle, walks WordprocessingML for paragraphs with run-level bold, italic, fonts, and colors, and catalogs tables, headers, and footers so you can map placeholders before generating filled documents. The workflow fits indie founders shipping proposals, compliance packs, or branded reports where the agent must understand template structure—not guess from PDF exports. Running under Deno with explicit read permissions keeps analysis sandboxed compared to opaque desktop macros. After analysis, you can plan content replacement, validate that every merge field exists, and hand results to a separate fill or PDF pipeline. It is analysis-first: use it when the bottleneck is understanding what is inside a template, not when you only need a one-off manual edit in Word.692installs11SummarizationSummarization is a journey-wide agent skill that helps you pick and execute the right kind of summary instead of defaulting to a generic bullet dump. Before condensing, it walks through purpose (decisions, background, research, quick understanding, recall), audience expertise, desired scope from one sentence to multi-page, what to emphasize, and structural format. That framing matters for solo builders who reuse the same source material for an investor paragraph, a technical appendix, and a social post. Install it when agents are asked to summarize, create a TLDR, condense content, or draft executive summaries across research, product docs, support macros, or launch copy. It pairs with documentation and planning skills when you need consistent voice and length without re-reading entire threads or PDFs.638installs12Electron Best Practiceselectron-best-practices documents a production-ready electron-vite configuration for solo builders shipping cross-platform desktop apps with React. It replaces separate webpack or rollup setups with one defineConfig that wires main, preload, and renderer builds, externalizes Node dependencies where required, and aligns entry paths for TypeScript main code, preload bridges, and a renderer rooted at src/renderer. Environment-aware options turn terser minification and sourcemaps on or off by NODE_ENV, and a @shared alias keeps types and constants consistent across isolated contexts. Invoke it when you are past the prototype stage and need IPC-safe structure, fast HMR during development, and predictable production bundles. The skill is narrowly focused on build-time configuration rather than distribution, code signing, or auto-update pipelines.635installs13System Designsystem-design is an assistive diagnostic skill that helps solo developers move from validated needs to architecture without defaulting to buzzword stacks or vague "figure it out while coding" plans. It is for indie builders in Claude-compatible agents when integration risk, scaling assumptions, or unclear boundaries would otherwise become late rewrites. The workflow centers on making trade-offs explicit—what you optimize for, what you defer, and which constraints actually drive structure—and on recognizing State SD0, where requirements are still fuzzy and the right move is to step back to requirements analysis rather than drawing boxes. It spans Validate when you scope the solution shape and Build when you refine components and APIs. It does not replace a full enterprise architecture review; it keeps decisions legible for one-person teams so the first implementation pass matches the problem you already validated.561installs14WorldbuildingWorldbuilding is a MIT-licensed diagnostic agent skill for fiction authors and indie game writers who need settings that feel historically grounded, not painted backdrops. Invoke it when technology or magic has not reshaped society, institutions read as author-designed rather than evolved, economies do not hold up, or non-human species feel like humans in costume. The skill does not generate worlds from scratch—that is systemic-worldbuilding—and it explicitly defers character-focused and plot-structure issues to character-arc and scene-sequencing. Solo builders shipping narrative games, serial fiction, or IP bibles use it mid-draft to name what is missing and get concrete interventions instead of vague “add more lore” advice. It works across genres and is typed as diagnostic mode in metadata, optimized for agent sessions where you paste setting notes or chapters for critique.554installs15Novel RevisionNovel Revision is a journey-wide agent skill for writers who edit novels and hit cascade failures—fixing dialogue exposes plot holes, restructuring chapters undermines theme, tightening prose breaks pacing. The skill encodes multi-level change management across thematic, structural, and manuscript layers so agents help you plan edits, trace downstream impact, and avoid endless fix-on-fix cycles. Solo builders shipping fiction products, newsletters serialized as chapters, or narrative-heavy games benefit whenever creative work spans more than a single pass. Invoke it when revisions at one abstraction level keep breaking others, or when you need diagnostic framing before large rewrites. It complements planning and review habits elsewhere in the Prism journey but stays anchored in long-form fiction craft rather than codebase refactors.545installs16Story AnalysisStory Analysis is an agent skill that walks you through a fiction diagnostic after you finish a draft. Solo builders and indie authors who ship serial fiction, newsletters with narrative arcs, or game lore can invoke it instead of vague “does this work?” chat feedback. The skill splits work into Short Story mode for standalone pieces and Chapter mode for sections inside a larger work, each with focused evaluation blocks such as narrative foundation and character construction. You answer concrete questions—for example whether the core conflict fits in one sentence and whether scope matches word count—so revisions target structure and stakes rather than line edits only. It is evaluative, not generative: it does not rewrite prose; it surfaces strengths, weaknesses, and improvement opportunities aligned with whether the draft achieves its stated narrative goals. Pair it with your revision pass before beta readers, serialization, or launch.535installs17Task DecompositionTask Decomposition is an agent skill that turns overwhelming development work into manageable units sized for human cognitive limits, clear boundaries, and minimal blocking. Solo and indie builders install it when a feature or epic cannot be estimated, the starting point is unclear, dependencies stall progress, or scope keeps expanding. The skill is assistive methodology—not implementation—centering on the principle that the goal is the right tasks: small enough to understand, large enough to deliver value, and independent enough to avoid queues. It walks through limits, slicing patterns, and when not to decompose further. Pair it during Validate when scoping MVPs and again in Build/PM when backlogs swell mid-sprint. It deliberately does not replace system-design for architecture choices or hands-on coding; it produces a clearer work breakdown you can track (e.g., with github-agile) before agents or humans execute.525installs18Story Idea Generatorstory-idea-generator is a generative agent skill for fiction and narrative-led products. It rejects “sci-fi as genre” in favor of emotional genres—wonder, horror, mystery, drama—and then selects setting modules that amplify that experience. Solo builders writing games, interactive fiction, or long-form content can use it when starting a project, when ideas feel thin, or when they want plot and character choices traced back to a deliberate reader feeling. The skill documents a modular system: a core elemental-genre table plus separate setting modules for science fiction, urban fantasy, epic fantasy, and historical fiction. That structure keeps brainstorming consistent without locking you into one world-building template. It pairs well with later validate-phase scoping when you turn a shortlisted concept into a prototype outline or pitch, but it does not replace market validation for commercial SaaS. Use it whenever creative direction should be driven by impact, not by picking a skin first.517installs