claude-skills

This skill should be used for elimination-style research where the user wants to choose from a shortlist of products, tools, services, vendors, or other options using explicit criteria, numeric evidence, tournament-style comparison, source/domain classification, image-supported consumer reports, raw data tables, and ownership-cost estimates for replaceable parts. Use this skill whenever the user asks to compare options, buy something, shortlist candidates, rank alternatives, generate a "don't make me think" report, or produce a full audit report with raw numeric data.

This skill should be used to run a formal heuristic evaluation of a design artifact against Jakob Nielsen's 10 usability heuristics, producing an evidence-backed, severity-scored report. Use it when the user wants a "heuristic evaluation", "usability review", "Nielsen heuristics check", "UX heuristic audit", or asks whether a screenshot, live URL, HTML page, codebase UI, interface description, or JTBD/spec document holds up against usability principles. Accepts five input types (screenshot/image, live URL, codebase/HTML, interface description, JTBD/spec doc) and adapts its rigor and output honestly to what is actually observable. Can render the report as plain markdown (default) or as a Tufte-style HTML report, and can export findings above a severity threshold as Linear or Beads (bd) tasks after confirmation.

This skill should be used when editing, translating, or reviewing an Astro-style i18n string corpus (files of the form export default { en: {...}, ru: {...} } under src/i18n/strings), or when the user wants to fill in missing translations, audit coverage, accept/review translations, propagate an edit across duplicate strings, get translation candidates, bulk-edit UI microcopy, or open a visual/keyboard translation editor. Drives the standalone i18n Studio tool at ~/ai_projects/i18n-studio (AST-safe minimal-diff saves via ts-morph, acceptance review state, duplicate propagation, hot-reload, and Claude translation suggestions).

This skill should be used when taking a single software feature from intent to shipped as a solo developer — goal-first, TDD, deterministic verification, evidence only where it earns its keep, and human judgment at the two moments that matter (goal approval, merge). Trigger when the user says "let's build feature X", "ship this feature", "run this through the factory", "write a goal contract", "feature-factory", or wants a disciplined intent→merge loop that resists process bloat. NOT for whole-product planning, multi-feature roadmaps, or autonomous multi-agent swarms.

A Theory-of-Constraints diagnostic for deciding what to automate with AI agents. Before building any automation, skill, Goal, loop, or schedule, it walks Goldratt's Five Focusing Steps over the user's work system to find the real bottleneck, then recommends the single highest-leverage automation aimed at the constraint plus a what-NOT-to-automate list. Use when the user asks "what should I automate", "where do I point my agents", "prioritize my automation backlog", "which workflow should I agentify", "is this worth building", "find my bottleneck", "what's the highest-leverage thing", when they are about to build a Claude Code skill/Goal/loop/schedule and aren't sure it matters, or during a review of their automations. Guards against the common failure of automating busywork (a local optimum) instead of the constraint.

This skill should be used to set up, validate, resolve, and export design tokens following the DTCG (Design Tokens Community Group) Format Module 2025.10 standard. Use when the user wants to define a design token set globally or per project, compile tokens to CSS variables, layer a project's tokens over a global brand base, or produce an on-brand context file for other generation skills. Triggers on "set up design tokens", "create a token set", "compile tokens to CSS", "design system variables", "brand tokens".

Controls the cmux macOS terminal app (Ghostty-based, AI-agent-aware) via its CLI socket API. Use this skill to: open browser preview panes when a dev server starts or an HTML file is created; spin up named workspaces for parallel subagents; show live sidebar progress/status during long tasks; open utility splits on demand (logs, URLs, terminals) in any direction. Also handles `/cmux demo` — an interactive walkthrough with 3 screens of options that runs personalized live demos. Trigger when: a server starts and needs previewing, an HTML/URL artifact is created, parallel subagents are dispatched, a long task needs visible progress, the user asks to open/show something in a split or tab, or the user invokes `/cmux demo`.

This skill should be used when importing, listing, or exporting Granola meeting recordings and transcripts. Queries Granola's Personal API to list meetings, extract transcripts, and export to Obsidian notes in Fathom-compatible format.

End-to-end pipeline for releasing an iOS / watchOS app to TestFlight and the App Store. Use when the user wants to publish, ship, or release an iOS/watchOS app, get a build onto TestFlight, archive and upload via xcodebuild, deploy a CloudKit schema to Production, set App Privacy or export compliance, mint a Distribution certificate, or work through App Store Connect / Apple Developer portal steps. Triggers on 'continue publishing', 'ship the app to TestFlight', 'release the iOS app', 'upload a build', 'deploy CloudKit to production', 'App Privacy labels', 'distribution signing', or any cryptic Apple upload error (cloud signing, 90057 missing CFBundleShortVersionString, 90474 orientation). macOS and Xcode only.

Run quick, offline, private LLM tasks on local models via llama.cpp, reusing models already downloaded by Ollama. Use for cheap/bulk text work (summarize, classify, extract JSON, anonymize PII, translate, proofread, keywords), local embeddings, and offline image description — and prefer it over a cloud API whenever a task is privacy-sensitive, must run offline, is high-volume/low-stakes, or just needs a fast throwaway answer. Provides an `lm` CLI wrapper plus an OpenAI-compatible local server.

Scan and clean macOS caches, package-manager data, crash dumps, and app caches to reclaim disk space. Deterministic — a config registry (targets.json) plus two scripts (survey.py read-only, clean.py executor) do all the measuring and deleting; the agent only relays a compressed summary and makes the few human-judgment calls. IMPORTANT — use this skill whenever the user's request on macOS involves: freeing disk space, cleaning/clearing caches, "disk is full", "clean up my Mac", "free up space", "what's eating my disk", "running low on disk", needing space for an install, or any low-storage complaint. Covers the whole workflow survey → choose → clean → empty Trash.

Interactively prepare a code repository for publication — LICENSE, NOTICE, AUTHORSHIP, README sections, package metadata, .gitignore, community docs (CONTRIBUTING/CODE_OF_CONDUCT/SECURITY/CHANGELOG), .github templates (issues/PR/CI/dependabot), a promo "more from the author" block, and conditional EU/Germany legal compliance (CRA, AI Act, GDPR, Impressum, product liability). Includes a dedicated authorship wizard that documents human decisions, judgment, direction, and art direction with jurisdiction-aware legal framing. Use when setting up a new repo for release, adding missing legal/meta/community files, writing an AUTHORSHIP record for AI-assisted work, or asking "prepare this repo for GitHub / open-sourcing".

Evidence-based CBT and DBT intervention skills — guided thought records, opposite action, DEAR MAN roleplay, crisis skills with optional HRV biofeedback. Configurable therapeutic pushback. Triggers on "/cbt", "/dbt", "/thought-record", "/record", "/opposite", "/opposite-action", "I need to work through something", "help me with a thought", "cognitive distortion", "I'm spiraling", "can we do a thought record".

Audit a white paper or long-form technical document against a research-grounded best-practices checklist. Two lanes — deterministic script checks (readability, undefined acronyms, structure blocks, broken links) plus an LLM-judge review (overclaims, inconsistent numbers, buried lede, limitations honesty, audience fit). Produces a prioritized P0–P2 findings report by default; applies fixes on explicit request (TDD for any code changes). Use when the user says "audit this white paper", "review my paper against best practices", "check this doc for overclaims", "whitepaper QA", "is this paper ready to publish", or wants prioritized recommendations on a technical document.

Scaffold a new Tauri v2 project with the cenno/cull house conventions — delegates boilerplate to `npm create tauri-app`, then layers an opinionated core plus opt-in modules (CLI+MCP, SQLite, tray/updater, release/preflight, Swift sidecar). Use when the user wants to start a new Tauri desktop app, "init a tauri project", or "scaffold a tauri app".

Scaffold a standalone SwiftUI app via XcodeGen — pick macOS, iOS, or iOS+watchOS; ships house conventions, a Swift Testing target, swiftformat/swiftlint, optional CloudKit/CI/Release/Push modules, and optional JTBD product context. Use to "init an xcode project", "scaffold a swiftui app", "new mac/ios app".

This skill should be used when analyzing Wispr Flow voice dictation history for self-reflection, work patterns, mental health insights, or productivity analytics AND when managing the Wispr Flow dictionary (adding terms, fixing mishears, exporting/importing, suggesting improvements). Triggered by requests like "/wispr-analytics", "analyze my dictations", "what did I dictate today", "wispr reflection", "add to wispr dictionary", "improve dictation", "wispr suggest", "export wispr dictionary", or any request to review voice dictation patterns or manage dictation quality.

This skill should be used when designing, running, validating, or auditing statistical experiments on personal or observational time-series data (health metrics, speech/text corpora, behavioral logs, diaries, n-of-1 self-tracking). It enforces pre-registration, exact permutation tests, FDR discipline, data-validation gates, adversarial code review, and cross-validation with external models. Triggers on "design an experiment", "test this hypothesis on my data", "is this correlation real", "audit these findings", "pre-register", "validate this dataset", or any n-of-1 / quantified-self analysis request.

Create Tufte-inspired data reports and infographic dashboards as standalone HTML files. Uses EB Garamond for text, Monaspace Argon for numbers, Chart.js for interactive charts, and inline SVG sparklines. Produces publication-quality reports with 2-column narrative+data layouts, status dashboards, scroll animations, and responsive mobile support. Use this skill whenever the user wants to create a data report, activity dashboard, infographic, personal analytics page, health tracker visualization, or any document that combines narrative text with interactive charts and tables. Also triggers for "make a report like Tufte", "create an infographic", "build a dashboard", "visualize my data", or requests for beautiful data-driven documents.

Build and verify a PII gold set with HUMAN annotators (first-class). Launch the browser annotator, label spans per the codebook, export per-annotator label files, then compute inter-annotator agreement (Cohen's/Fleiss' kappa) and draft an adjudicated gold. Use when the user says "annotate PII", "label this transcript", "build a gold set", "inter-annotator agreement", "review annotations", "adjudicate labels", or wants to measure/defend a de-identification gold standard. Local-only: synthetic or consented data only; annotators' names and transcript text stay on the machine — only labels/stats are collected, nothing PII is re-shared.

De-identify a session transcript (file or folder) by redacting PII LOCALLY before any sharing or cloud use. Produces a redacted GREEN copy with unique reserved-sentinel placeholders ([CONFIDE_PERSON_0001], [CONFIDE_EMAIL_0001], [CONFIDE_DATE_0002]...) plus a counts-only stats summary, and a local secret <name>.map.json (0600, gitignored) that enables confide:rehydrate to restore real values after a cloud analysis. Use when the user says "anonymize this transcript", "redact PII", "de-identify session", "make safe to share", "strip personal data", "anonymize notes before sending to an LLM", or points at a transcript/folder that should be scrubbed. Local-only by default — raw text never leaves the machine; the map is the only artifact with originals and stays local; nothing printed is PII; human review is still required before sharing.

Run a corpus-scale, STATS-ONLY PII audit over a folder of session transcripts LOCALLY and produce an aggregate report — counts by type and by layer, the per-session redaction-rate distribution, document lengths, and a coarse residual proxy. Use when the user says "audit my sessions", "scan folder for PII", "how much PII across these transcripts", "PII stats for my corpus", "is my redaction holding at scale", or points at a directory of transcripts and asks how much personal data it contains. Fully local — raw text never leaves the machine; the report carries ZERO PII values, transcript substrings, or filenames (only anonymized own-NN ids and counts), so the aggregates are safe to surface. Run it on a RED (raw) corpus to size the PII, or on a GREEN (already-redacted) corpus to check residual leakage.

Put the real values back into an analysis that was produced from GREEN (placeholder) text — LOCALLY, using the user's own reversible map. Completes the confide round-trip (redact -> cloud-analyze the green -> rehydrate locally). Use when the user says "rehydrate", "restore real names", "unmask the analysis", "put the names back", "de-redact this output", "reverse the placeholders", or hands you an analysis full of [CONFIDE_PERSON_0001]/[CONFIDE_DATE_0002] plus a *.map.json. Runs only on the user's own map; the map never leaves the machine; nothing fetched or transmitted. Prints counts only — never echoes restored PII. Warns on placeholders not in the map (possible LLM hallucination).

Set up and verify the CONFIDE THREE LOCKS for storing RED (real, identifiable) session data at rest — device FileVault, a dedicated encrypted store, and per-file sops/age encryption. Use when the user says "set up confide vault", "encrypt my session data", "three locks", "secure store for transcripts", "sops/age for RED data", or asks how to store real therapy/coaching transcripts safely. NON-DESTRUCTIVE: it CHECKS each lock's status and prints the EXACT command to fix any gap; it never moves, deletes, or encrypts data, and never runs `fdesetup enable`/`hdiutil`/`age-keygen` without an explicit flag and your confirmation. Probes are read-only (`fdesetup status`, which sops/age, key path).

Build a self-contained interactive HTML that lets you SEE de-identification and restoration — original ↔ redacted ↔ rehydrated, with color-coded PII spans and All/None/Selected toggles. Use when the user says "show me what was redacted", "visualize the de-id", "compare original and redacted", "highlight the PII", "view redaction diff", "see what rehydrate restored", or wants to inspect exactly which spans each type/layer removes. LOCAL only — the HTML embeds REAL values (like a vault artifact): written locally, gitignored, banner-marked private, never shipped or committed.

Residual re-identification RISK CHECK on text you have ALREADY redacted (defensive, dual-use). Use when the user asks to "check residual re-id risk", "red-team my redaction", "what can an attacker still infer", "is this safe to share", or assess "re-identification risk" after anonymizing. Re-runs the CONFIDE detectors on the redacted output to surface surviving identifiers (singling-out), checks multiple files for linkability, and optionally probes a local model for still-inferable attribute CATEGORIES (inference) — mapped to GDPR Art-29. Reports risk categories/counts only, never a re-identification recipe. Pairs with confide:anon (run AFTER redacting).

Set up, install, and configure CONFIDE local de-identification — installs Python deps (natasha, scrubadub, phonenumbers, pymorphy2), ensures Ollama + pulls the default qwen2.5:3b model, detects optional llama.cpp, and writes the optimal-default config so confide:anon and confide:red work with zero further config. Everything is local-first; raw text never leaves the machine. Use when the user says "set up confide", "install confide", "configure confide de-id", "confide setup", or "get confide ready".

Generate and edit images using OpenAI's GPT Image 2 API. Interactive skill that guides users through image creation with style presets, cost-aware draft/final workflow, thinking mode, carousels, and photo editing. This skill should be used when the user requests image generation via OpenAI/GPT Image 2, wants to create social media carousels, edit photos into artistic styles, or needs images with readable text (infographics, diagrams, posters).

Cut a software release and maintain a tiered compatibility policy. Use when the user wants to release, ship a version, bump the version, tag a release, write a changelog, or update COMPATIBILITY. Config-driven via release.config.json; bumps version files, runs a readiness gate, updates COMPATIBILITY.md tiers and deprecations, tags (→ release workflow), and reports closed issues. Teaches the underlying standards as it runs.

Use when automating a real, logged-in Chrome WITHOUT disturbing the user's open tabs — e.g. authenticated sessions like ChatGPT, LinkedIn, or any site needing a persistent login. Launches or reuses a dedicated debug Chrome instance per named profile that coexists with the user's main browser; it never quits or kills any browser (unlike real-browser, which closes Chrome Beta's tabs). Trigger on "automate this site without closing my tabs", "use my logged-in session", "open ChatGPT/LinkedIn in automation", or when a browser task must preserve the user's existing windows. macOS, requires the agent-browser CLI.

End-to-end pipeline for publishing Claude Code lab meetings. Accepts optional args: date (YYYYMMDD, "yesterday", "today") and lab number (e.g. "04"). Examples: "yesterday 04", "20260420 05", "04" (today, lab 04), "" (today, auto-detect lab).

Add agent-friendly --json NDJSON output to Python CLI scripts, or scaffold a complete cli_utils package for a project. Use this skill when the user wants to make scripts machine-readable for AI agents, add --json flags, convert print statements to structured JSON, build a CLI helper library, create an open-source CLI-for-agents package, add structured logging, or make CLI output machine-readable. Also use when the user mentions NDJSON, structured CLI output, agent-friendly CLI, non-interactive scripts, or JSON I/O for automation.

Constructive, evidence-based dialogue mode that avoids sycophancy. This skill should be used when the user wants balanced multi-perspective analysis, critical feedback, or rigorous challenge of their ideas. Triggers on "/balanced" or requests for honest/critical/balanced feedback. Supports passive, interactive, tldr, steelman, and decision modes.

Applies Agency brand colors and typography to artifacts including presentations, SVG graphics, documents, and web interfaces. This skill should be used when brand colors, visual formatting, neobrutalism style, or Agency design standards apply. Keywords - branding, corporate identity, visual identity, styling, brand colors, typography, visual formatting, visual design, neobrutalism.

Query browsing history from all synced devices (iPhone, Mac, iPad, desktop). Supports natural language queries for filtering by date, device, domain, and keywords. Uses LLM classification for content categories. Can output to stdout or save as markdown/JSON to Obsidian vault.

Query Chrome browsing history with natural language. Filter by date range, article type, keywords, and specific sites.

Generate interactive AI transformation context-builder prompts for consulting clients. Use when creating structured discovery session prompts that guide a company through context gathering about their business, pain points, tech stack, and AI opportunities. Produces a resumable, multi-section prompt with Express/Deep Dive modes.

Generate structured decision-making tools — step-by-step guides, bias checkers, scenario explorers, and interactive dashboards. Use when facing significant choices requiring systematic analysis. Supports multiple cognitive styles and output formats.

This skill should be used when conducting comprehensive research on any topic using the OpenAI Deep Research API. It automates prompt enhancement through interactive clarifying questions, saves research parameters, and executes deep research with web search capabilities. Use when the user asks for in-depth analysis, investigation, research summaries, or topic exploration.

Evidence-based health research using tiered trusted sources with GRADE-inspired evidence ratings. Integrates Apple Health data for personalized context. Use when user asks health, nutrition, exercise, sleep, or wellness questions.