AI Agent Workflow Collection

Create a scoped action brief from the current conversation. Use when a conversation produces something actionable that needs to be captured as a brief. Invoked by the cofounder agent or other skills when actions emerge.

Use this when the user wants BosskuAI to behave like an expert AI cofounder across product, engineering, UI/UX, security, SEO/GEO, GTM, marketing, sales, prioritization, and next-step decision making.

Language-agnostic code commenting and docblock conventions. Trigger on every source-code Edit/Write in any language (JS/TS, Python, PHP, Ruby, Go, Rust, Java, Kotlin, Swift, C/C++, C, Elixir, Lua, Shell, etc.) — even when the user has not mentioned comments, docblocks, JSDoc/TSDoc/PHPDoc, docstrings, or documentation. Governs every comment authored or modified, every new function/method/class/component (which must get a compliant docblock), every inline comment (which must stay terse — no prose walls, no business-logic essays, no "step 1 / step 2" narration in the code path), and any non-compliant docblock encountered in the file being edited (which must be fixed in place — no need to ask first, except when the existing docblock is unusually long and clearly intentional, in which case leave it alone). Apply to all code authoring across all languages, every time source code is touched. When in doubt whether this skill applies to a code edit, invoke it. Skip only for non-code files (JSON/YAML/TOML configs, markdown, lockfiles) or trivial one-character edits.

Use whenever the user has a plan, spec, RFC, ticket, design doc, or rough idea they want stress-tested before code is written. Triggers on "drill me", "interview me", "interrogate this plan", "stress test this", "what am I missing", "make sure we're aligned", "poke holes in this", "drill into this", "before I build this..." with an idea attached, or any time non-trivial work is about to start and pre-implementation alignment matters. Walks the design tree one branch at a time, asks one opinionated question per turn with a recommended answer, and explores the codebase instead of asking when the answer is already there. Continues until shared understanding is reached. Use this even when the user doesn't explicitly invoke the skill, as long as the situation matches — pre-implementation drilling is undertriggered by default.

Complete reference for implementing Driver.js (driver.js) product tours, element highlighting, onboarding flows, and contextual help overlays. Use this skill whenever the user mentions "driver.js", "driverjs", "product tour", "onboarding tour", "guided tour", "element highlighting", "walkthrough", "feature introduction", "spotlight overlay", or asks to build step-by-step user guides, interactive walkthroughs, or highlight-and-explain UI flows in web applications. Also trigger when the user wants to add tooltips that walk users through a page, create async/dynamic tours, or implement "turn off the lights" focus effects. Do NOT use for unrelated tooltip libraries (Tippy.js, Floating UI) or for general CSS overlay questions that don't involve guided tours.

Generates professional Lottie animations from static SVGs. Replaces After Effects for motion graphics. Use when the user asks to: animate logo, create lottie, svg animation, motion graphics, wiggle animation, bounce effect, rotate animation, pulse effect, entrance animation, loading animation, loop animation, icon animation, character animation, morphing, path drawing, trim path, walking animation, run cycle, walk cycle, frame-by-frame animation, sprite animation. Supports advanced bezier curves, shape modifiers, parenting, mattes, morphing, character rigs, and professional frame-by-frame animation techniques.

Post a structured investigation/fix comment on the active GitHub PR capturing findings, root cause, proposed fix, and a progress checklist — and later update progress or mark the comment resolved. Proactively invoke whenever an investigation or debugging session on a PR branch has produced a concrete root cause plus a fix plan (even if the user hasn't asked yet) — the skill's first step is to propose posting the comment via AskUserQuestion, so firing it speculatively is cheap. Also invoke on explicit triggers like /make-issue-comment, "post this as a PR comment", "file these findings on the PR", "update the PR comment", or "mark that comment as resolved".

Extract text from images and scanned PDFs using OCR. Supports 100+ languages, table detection, structured output (markdown/JSON), and batch processing.

Offer to open markdown documents in mdreader after writing them. Trigger this skill whenever you write or create a markdown file that is a plan, spec, design document, architecture doc, implementation guide, proposal, RFC, or any substantive document meant to be read (not just a config snippet or small edit). Also trigger when the user asks to "open", "view", or "preview" a markdown file. Do NOT trigger for trivial edits to existing files (like fixing a typo in a README), code comments, or CLAUDE.md updates.

TanStack Query (React Query) patterns for server state management, caching, mutations, optimistic updates, and infinite queries.

Reshuffle tasks and priorities when stuck or drifting. Identifies the critical path and parks non-essential work to protect deadlines.

Attempt-compare-gap loop for deliberate…

Deterministic SVG generation, validation, and rendering. Use for icons, diagrams, charts, UI mockups, or technical drawings requiring structural correctness and cross-viewer compatibility.

Summarize recent git work for time registration. Use when the user asks what they worked on, needs a time log, wants a work summary, asks about today's/yesterday's/last week's work, or invokes /time-registration:time-registration. Accepts an optional time period argument (defaults to today).

Control the user's real macOS screen via the agent-vision CLI — session management, element targeting, and UI interaction on live windows. Triggers: user says "look at my screen", "use agent-vision", "the app/browser/simulator is open", "take a screenshot of my screen", "fill this form" (when app is already open), "check the UI", "watch the browser", "navigate to" (in an open app), "scroll through", "click on" (in a visible window), "I have X open", visual QA of running applications, iOS Simulator or Android emulator interaction, before/after visual comparison of live UI, or any task requiring real screen capture and control. NOT for: headless browser testing, Playwright/Puppeteer scripts, code-only reviews, file-based screenshots, or building screen capture features.