daymade/claude-code-skills
24 skills16.2k installs27.8k starsGitHub
Install
npx skills add https://github.com/daymade/claude-code-skillsSkills in this repo
1Twitter Readertwitter-reader is a small Python workflow that turns an X/Twitter article URL into a local Markdown clipping with downloaded images. Solo builders use it when a competitor thread, launch post, or founder essay lives on X and they want the same material in their repo or second brain without screenshots or broken embeds. The script shells out to twitter-cli through uv for structured fetch, parses YAML output, materializes attachments, and writes a portable note. When twitter-cli is not enough, an optional Jina API path adds reader-style text extraction. It fits indie research habits: archive what you cite, grep it later, and feed summaries to your agent. Expect dependencies on uv, twitter-cli, Python 3, network access, and optionally a Jina key—not a hosted SaaS, but a repeatable clip command you own.1.7kinstalls2Prompt Optimizerprompt-optimizer is a reference skill for solo builders and small teams who need agent and system prompts grounded in testable requirements instead of prose wishes. It teaches Advanced Prompt Optimization through EARS transformation patterns aimed at complex domains: multiple stakeholders, non-functional constraints, and layered conditionals. You reach for it when a feature touches different roles with conflicting permissions, when access rules depend on more than one factor, or when you want shall-statements an engineer or coding agent can implement without reinterpretation. The readme walks through role-based, hierarchical, and context-dependent access templates you can adapt to SaaS admin panels, internal tools, or autonomous agent guardrails. It complements narrative product docs by forcing explicit if/when structure, which reduces ambiguity in later build and ship review cycles. It does not replace discovery interviews or implementation plans; it sharpens the requirement language those artifacts depend on. Builders using Claude Code, Cursor, or Codex benefit when specs must double as durable prompt constraints for downstream agents.1kinstalls3Ppt CreatorPPT Creator is an agent skill for solo builders who need investor pitches, product demos, business reviews, or training decks without living in PowerPoint. It starts from a topic or scattered source material, optionally runs a minimal 10-question intake when details are thin, then follows a documented workflow to produce an outline, slide copy, data-driven charts, and speaker notes. Content is shaped with persuasive frameworks such as the Pyramid Principle and assertion-evidence pairs so slides read as claims backed by proof, not bullet soup. Before handoff, the agent scores the deck against an internal rubric and automatically refines up to two iterations if the score stays under 75. Use it whenever you want “make me a deck” quality with consistent structure rather than one-off chat slides.1kinstalls4Qa Expertqa-expert gives solo builders a consistent test-case authoring template so QA artifacts are reproducible whether a human or an agent executes them. Each case documents priority, test type, prerequisites, exact steps, expected outcomes with checkmark-style success criteria, and explicit pass versus fail rules—reducing ambiguous “it seems fine” reviews before ship. The included CLI install example shows how to structure integration coverage for agent skills themselves. Use it heavily in Ship while expanding regression suites, in Validate when proving a prototype path, and in Operate when turning incidents into new regression cases. It does not replace automated test runners; it standardizes the human-readable spec those runners should implement.904installs5Macos Cleanermacos-cleaner is a reference skill for solo and indie builders who live on macOS and need disciplined disk cleanup without nuking real data. It walks through what each cache directory holds—browser and app caches, thumbnails, fonts, and shared system caches—and labels targets as safe to delete when the goal is space, not preservation. The skill emphasizes operational tradeoffs: deleting caches does not lose user documents, but you should expect cold-start slowness until apps rebuild their stores. Package-manager sections call out Homebrew’s cache path and the brew cleanup workflow so npm-heavy and brew-heavy stacks both get attention. Use it when your SSD is full, Docker or Xcode leftovers are suspected, or you want an agent-guided checklist instead of forum folklore. It is host maintenance, not app deployment—best invoked from Operate when monitoring local resources or before cloning large monorepos.737installs6Deep ResearchDeep Research is a multi-phase agent skill for rigorous, citation-first research reports with an automated counter-review team before delivery. Indie builders and operators using Claude Code invoke it when chat summaries are insufficient and every claim, date, and number must map to sources. The workflow includes a completeness checklist covering section order, evidence tables, limitations, and terminology, then P6 parallel agents validate claims, source diversity, recency (AS_OF), and internal contradictions. It suits early discovery, investment or product validation memos, and any grow or launch narrative that must survive scrutiny. Outputs are structured drafts and reviewer reports—not raw search dumps—so you can hand results to planning, docs, or distribution with traceable evidence.683installs7Ui Designerui-designer is an agent skill that pairs a SaaS-founder-style product consultant with a professional UX/UI style-guide writer. Solo builders invoke it when they have an application idea or MVP sketch and need a repeatable document instead of ad-hoc chat fragments. The first half walks through elevator pitch, problem statement, target audience, unique selling proposition, target platforms, and a features list organized by category with checkbox-style user stories and nested acceptance criteria. It also forces explicit UX/UI consideration per screen or flow, including how different states look, animate, and layer information. The second half produces a visual style guide from attached reference images, wrapped in a designer-style pondering block before outputting color roles and typography-like structure. The skill is meant for iterative co-editing: each user reply is merged back into a full refreshed plan. It suits indie founders validating scope and visual direction before implementation, especially for multi-platform SaaS or consumer apps.650installs8I18n Experti18n-expert packages a procedural audit for solo and indie builders who ship multilingual web apps with i18next-style APIs. Instead of manually grep-ing for broken translations after a release, you point the included script at your src tree and one or more locale JSON files; it extracts every translation key referenced in code and reports gaps against what is actually defined on disk. The workflow fits teams using React, Vue, or similar stacks where strings live in code and JSON locale bundles. It is intentionally narrow: it does not translate copy or manage TMS workflows—it validates consistency so agents and humans do not merge PRs with silent missing-key fallbacks. Use it during feature work, before cutover to a new language, or in CI as a gate. Complexity is beginner-friendly because the interface is a single Python command with --src and --locale flags.626installs9Fact Checkerfact-checker gives solo builders and small teams a repeatable way to stop stale specs and wrong numbers from shipping in markdown, READMEs, or internal docs. The agent scans for verifiable claims—context windows, pricing, release dates, API capabilities, benchmarks—and ignores subjective recommendations and how-to prose. It follows a five-step workflow: identify claims, search official and authoritative sources, compare, generate a correction report, then apply edits only with your confirmation. Use it when someone asks to fact-check, verify accuracy, validate a section, or refresh outdated data, especially for AI model documentation where specs change quickly. It complements ad-hoc chat by forcing source-backed diffs instead of silent rewrites.604installs10Developing Ios AppsDeveloping iOS Apps is an agent skill for solo and indie builders shipping Apple-platform products with XcodeGen, SwiftUI, and Swift Package Manager instead of hand-maintained Xcode projects. It activates when you hit project.yml churn, SPM dependency or embed errors, real-device deployment blocks, or the long tail of signing and notarization failures that stall TestFlight and Mac distribution. The skill encodes critical warnings in table form—such as dynamic framework embedding, free Apple ID limits on command-line signing, and CI conditionals that leak secrets—so your coding agent proposes fixes aligned with how Apple tooling actually behaves. It also covers AVFoundation camera debugging and iOS version compatibility when runtime loads break in the field. Use it while building the app, hardening release pipelines, or unblocking a red CI job before you can ship a build to testers or the App Store.597installs11Skill CreatorSkill-creator (as documented in the Post-hoc Analyzer portion of the package) helps solo builders who maintain Claude Code skills turn subjective “which answer felt better?” judgments into structured learning. After a blind comparator picks side A or B, the analyzer reads both skills’ SKILL.md files, key referenced assets, and execution transcripts, then contrasts instruction clarity, script usage, and other structural cues against what the comparator valued. That closes the loop for indie skill authors who cannot run a full ML eval team but still want measurable iteration on procedural knowledge. Prism places it on Build → agent-tooling because its deliverable is a better skill definition—not a shipped product feature. Use it when you already run paired executions and have comparison_result_path style inputs; skip it when you have not defined a baseline skill or comparator yet. The workflow assumes explicit file paths for winner and loser artifacts and writes analysis JSON to a configured output_path for the next editing pass.593installs12Meeting Minutes TakerMeeting Minutes Taker is a checker-style agent skill for solo and indie builders who capture calls and need minutes that hold up as source-of-truth for specs. It walks a four-phase completeness review: structural coverage of topics, deep content checks for entities and numerical detail, quote verification for significant decisions, and an action-item audit against the transcript. The embedded checklist calls out common omission patterns across technical models, business logic, and process items, plus merge guidance when multiple draft versions exist. Use it after transcription or an initial draft in validate or build so prioritization, state machines, and rejected alternatives are not lost before you write plans or open issues. It does not replace live note-taking—it hardens what you already captured.587installs13Promptfoo EvaluationPromptfoo Evaluation is an agent skill reference for solo builders shipping LLM-powered features who need repeatable prompt tests without guessing YAML syntax. It catalogs how to wire Anthropic and OpenAI providers, run cost-free echo previews that surface substituted variables and few-shot layout, and structure multi-provider experiments with clear labels. The assertion section explains the Python evaluation context—prompt, vars, test case, config, provider response—so you can encode business rules beyond string contains checks. Echo mode is especially valuable when iterating agent skills: you validate rendering and case matrices before spending tokens on live models. Pair this skill when hardening Ship-phase quality gates or when Build-phase agent tooling needs a shared eval vocabulary. The bundled security scan metadata reminds you to treat eval configs like production secrets when API relays are involved.586installs14Mermaid ToolsMermaid-tools is a Claude Code skill package that documents how to generate Mermaid diagrams reliably on developer machines, with extra emphasis on WSL2 and headless Chrome. Solo builders use it when architecture or README content already contains Mermaid syntax but PNG or SVG exports fail or never get automated. The readme walks through installing mermaid-cli globally, verifying mmdc, installing Chrome and GTK-related libraries on Linux/WSL, and locating the skill’s extract-and-generate shell and Python scripts via CLAUDE_SKILL_DIR. It also covers environment variables, common failure modes, and a validation checklist so agents do not guess at one-off fixes. It is not a hosted diagram service; it is procedural knowledge plus local scripts for extraction and batch generation tied to your repo.578installs15Statusline Generatorstatusline-generator is an integration-oriented agent skill for solo builders who run Claude Code daily and want spend visible in the status bar. It explains how a statusline script should invoke ccusage to read conversation transcripts, format session and daily totals as fixed-precision currency strings, and stay fast enough for interactive use. The reference covers offline JSON queries, descending sort for the most recent session or day, and a deliberate caching layer so the UI does not block on pricing fetches. First paint may omit costs for a few seconds while a background refresh runs; older cache files bridge that gap. You install it when you already use or plan to use ccusage and want copy-paste shell patterns rather than guessing at flags and error handling. It does not replace ccusage installation or broader budgeting workflows—it focuses on reliable, low-noise statusline wiring for indie operators watching token burn.563installs16Skill ReviewerSkill Reviewer is an agent skill that acts as a quality gate for Claude Code skills you write or fork. Indie builders maintaining a personal skill library use it when a new SKILL.md is draft-complete but might fail discovery, misfire on triggers, or bury instructions in chatty prose. The workflow mirrors a formal checklist: validate name length and reserved words, ensure descriptions are third-person with concrete when-to-use clauses, confirm imperative infinitive instructions, and verify progressive disclosure splits heavy reference material from the core flow. It fits naturally after brainstorming or drafting a skill and before you ship the skill to teammates or a marketplace listing. The skill is meta— it improves other skills—so it stays in the agent-tooling lane rather than application feature work, though the same passes help during operate-phase maintenance when upstream templates change.561installs17Skills SearchSkills-search is an agent skill that wires your coding agent directly into the CCPM (Claude Code Plugin Manager) registry so you can discover, install, and manage community skills from inside a session. Solo builders use it when they know they need a capability—PDF handling, code review, Cloudflare troubleshooting—but do not want to hunt GitHub or paste shell commands manually. On activation it checks for `ccpm`, bootstraps the ecosystem in one step when needed, and executes the right CLI action via Bash while honoring allowed-tools constraints. It fits the build phase’s agent-tooling shelf and doubles as early journey discovery when you are still assembling your toolkit. The skill is opinionated about agent behavior: execute first, explain second, which keeps momentum for indie shipping workflows tied to Claude Code and compatible desktops configured through setup.542installs18Youtube DownloaderYouTube Downloader is an agent skill that encodes an internal standard operating procedure for reliable YouTube acquisition with yt-dlp, not a generic "paste a URL" shortcut. Solo builders and indie operators use it when automation hits YouTube's Proof of Origin requirements, zsh glob pitfalls on query strings, or proxy mismatches between the downloader and the token provider. The playbook walks through quoting URLs, activating proxies end-to-end, requesting browser cookies when bot challenges appear, starting PO token providers before download, and choosing the right client profile when cookies are present. It also documents failure signatures—403s, image-only streams, format unavailable errors, and unstable proxy fragments—and maps each to concrete retries. For Prism's audience shipping content pipelines, research archives, or media tooling inside Claude Code–style agents, this skill reduces wasted shell cycles and ambiguous agent guesses by turning tribal yt-dlp knowledge into procedural steps agents can follow consistently.536installs19Claude Code History Files FinderClaude Code History Files Finder helps solo builders who live in Claude Code dig through on-disk session history when something vanished from the chat UI or never got saved to the repo. The skill packages search and recovery scripts plus a reference for how session files are structured as JSONL. analyze_sessions.py supports finding relevant sessions and analyzing them; recover_content.py targets pulling usable text back out when you know roughly where it lived. It sits in Build as agent-tooling because that is when you are actively coding with the agent and most likely to need recovery, but the same workflow appears in Operate when debugging a bad session or auditing what the agent actually wrote. It is a filesystem-local integration, not a hosted backup service—your machine must still have the history files present.534installs20Github OpsGitHub Ops is a reference agent skill that teaches solo and indie builders how to drive GitHub through the GitHub CLI gh api command. Instead of memorizing REST paths or guessing query parameters, you install procedural knowledge for listing and filtering pull requests, managing issues, touching repository settings, triggering or inspecting Actions workflows, running search queries, and optionally using GraphQL when REST is awkward. Authentication is anchored on gh auth status and tokens from gh auth login, with explicit header guidance for the stable GitHub API version. The skill fits builders who ship from Claude Code, Cursor, or Codex and want repeatable automation—opening PRs, triaging issues, wiring CI checks, or scripting repo hygiene—without leaving the agent thread. It is reference-style integration knowledge, not a hosted MCP server; your agent still executes shell commands against gh. Use it whenever you would otherwise tab-switch to fragmented GitHub docs mid-task.532installs21Claude Md Progressive Disclosurerclaude-md-progressive-disclosurer is an agent skill that applies progressive disclosure to CLAUDE.md and companion reference files so solo builders maximize how much useful guidance fits in the context window without duplication or maintenance drift. It is for anyone using Claude Code (or similar) on a real repo where instructions have grown messy, the same policies appear in multiple files, or the model keeps missing stated rules. The workflow emphasizes information efficiency in Anthropic’s sense: the smallest high-signal token set that still produces compliant behavior. You reorganize Level 1 content—principles, command tables, hard bans with examples, error diagnosis, copy-paste patterns, directory maps, and pointer indexes—and move lengthy SOPs, edge cases, and historical decisions into Level 2 references loaded only when needed. Multiple intentional entry points support different mental models when debugging versus before editing code. The skill explicitly forbids treating shorter files as success metrics, aligning optimization with maintainability and rule adherence instead of arbitrary trimming.526installs22Docs Cleanerdocs-cleaner is an agent skill that walks you through section-by-section documentation value analysis instead of blind deletes. Solo and indie builders use it when a repo has grown overlapping READMEs, agent instruction files, and ad-hoc notes that duplicate each other or bury essential setup steps. The workflow classifies each section as keep, condense, or delete using explicit criteria—unique troubleshooting, one-time meeting notes, self-evident code comments, and superseded guides. Before proposing removals, the checklist forces you to map overlap, name unique value per file, and choose one canonical location, then plan updates to CLAUDE.md and README links. Outputs are a value analysis table and a consolidation plan you can hand to an agent for controlled edits. It fits mid-build doc passes and later operate-phase cleanup when onboarding friction returns.526installs23Excel AutomationExcel Automation is an agent skill for macOS builders who need real Excel files—not CSV substitutes. It covers three paths: generating polished reports and mockups with openpyxl, extracting sheets and fixing corrupted names from financial models that break normal parsers, and programmatically steering the Excel app window for repeatable captures or reviews. Solo founders use it when investor models, ops dashboards, or client deliverables must stay in native xlsx/xlsm format while the coding agent does the heavy lifting. The workflow assumes uv for Python deps, documents when zip/XML parsing beats openpyxl, and warns that AppleScript control can hang without explicit timeouts. It fits indie teams shipping internal tools, finance-adjacent SaaS, or agency reporting pipelines where Excel remains the contract format.511installs24Capture ScreenCapture Screen teaches agents a dependable macOS screenshot pipeline when marketing pages, README walkthroughs, or compliance evidence need pixels from real application windows—not stale mocks. Solo builders start by obtaining a numeric window ID through CoreGraphics APIs because title-only guessing fails under modern macOS security. They optionally drive the target app with AppleScript to scroll lists, zoom worksheets, or select cells, then call screencapture with -l to grab exactly that window. The skill pairs naturally with Excel or other GUI-heavy tools when Excel Automation prepares the view and this skill records it. It is procedural CLI knowledge rather than a hosted screenshot SaaS, which keeps indie workflows local and repeatable for multi-shot documentation series.510installs