sebastian-software
OrganizationPractice-built skills that give AI agents practical judgment for product, web, engineering, delivery, go-to-market, communication, and compliance work.
Categories
Indexed Skills (18)
codebase-improvement
Audit, diagnose, plan, and improve software codebases through repository reconnaissance, evidence-backed findings, competing-hypothesis investigation, leverage-based prioritization, executable plans, backlog reconciliation, and proportionate simplification. Use when asked why a defect or surprising behavior occurs, for a read-only root-cause report, to audit a repository, decide what to improve next, simplify code, create or review an implementation plan, reconcile stale work, or implement selected improvements. Do not use for PR-only review, dependency-update-only work, focused test, test-runner, or test-framework diagnosis, architecture-only decisions, or frontend-only diagnosis when a narrower skill applies.
consultant-profile
Create, audit, rewrite, localize, and improve consultant CVs, LinkedIn profiles, case studies, project lists, bios, and senior technology-consulting resumes. Use for positioning, profile completeness, career narratives, evidence-led project selection, cross-channel synchronization, and interview-driven discovery. For the profile as a customer-acquisition or pipeline surface, use linkedin-social-selling.
decision-records
Create, review, update, supersede, and audit Architecture Decision Records (ADRs) for durable project decisions. Use when a user asks for an ADR, decision log, architecture decision, design decision, communication or brand voice decision, or recorded rationale; when a cross-cutting technical, product, design, content, marketing, security, operational, or process choice must remain understandable across people, tools, and agent sessions; or when another skill identifies a decision whose undocumented rationale would otherwise drift. Preserve existing repository conventions and do not create skill-specific dot folders or private memory formats.
effective-web
Diagnose, design, build, review, and improve user-facing web experiences: marketing sites, content pages, landing pages, web apps, dashboards, and React components. Covers UI/UX direction, layout, typography, color, components, forms, tables, accessibility, motion, internationalization (i18n) implementation, interface copy, loading and error states, auth UX, CSS architecture, frontend SEO and AI search, AI-assisted interface design, browser performance, frontend-only surprising behavior, frontend testing, React architecture and component APIs, SVG graphics (icons, paths, animation, textures), print stylesheets, paged media, and web-to-print output. Not for: backend-only work, infrastructure or deployment, general-purpose writing, print production unrelated to HTML and CSS, or locale-specific punctuation, quotation, spacing, and language-level typographic rules (locale-typography skill).
effective-workflow
Coordinate multi-stage or mixed-domain software work from an unclear request through an authorized change, focused verification, and a review-ready handoff. Use when a task spans diagnosis, implementation, validation, review, or delivery and needs sequencing across repository-native workflows and first-party specialist skills. Prefer the matching specialist directly for an already narrow documentation, dependency, test, review, architecture, port, or frontend request.
linkedin-posts
Create LinkedIn posts, find LinkedIn content ideas, plan a LinkedIn content calendar, and tailor professional posts to an audience, voice, and goal. Use when the user mentions a LinkedIn post, LinkedIn content, LinkedIn idea, social media post for LinkedIn, post engagement, or content calendar for LinkedIn. Route engagement-to-pipeline diagnosis and acquisition strategy to linkedin-social-selling.
linkedin-social-selling
Plan, build, review, and improve an evidence-led B2B LinkedIn customer acquisition system. Use for LinkedIn positioning and offers, inbound profile strategy, target-account networking, welcome and outreach messages, social selling conversations, content systems, lead magnets, sales funnels, and measurement across the path from profile visit to qualified meeting. Use when the request spans multiple LinkedIn growth activities or connects content and relationships to pipeline; route a post-only writing request to linkedin-posts and field-level profile content or career evidence to consultant-profile.
locale-typography
Apply and review locale-appropriate typography in visible prose for German, English, French, Dutch, Luxembourgish, European Spanish (es-ES), Italian, Danish, Norwegian, Swedish, Polish, Czech, and Croatian. Use when writing, editing, localizing, proofreading, or implementing text in Markdown, HTML, JSX, templates, UI copy, documents, or print layouts where quotation marks, punctuation spacing, dashes, apostrophes, non-breaking spaces, numbers, dates, currency, or hyphenation must match a supported language and region.
metro-english
Rewrite text into relaxed US metropolitan team English: the kind of natural, direct, human voice used in Slack threads, GitHub or Linear issue comments, PR notes, async updates, and internal team communication. Use this skill whenever the user asks to make writing sound casual, less formal, less AI-generated, more human, more Slack-like, more startup/team-like, or closer to everyday New York, San Francisco, or Silicon Valley professional English. Also use it when the user provides German or stiff English and wants it turned into natural English for team communication.
port-codebases
Plan, execute, review, and verify behavior-preserving codebase ports across programming languages, runtimes, frameworks, platforms, storage engines, or major APIs. Use when asked to rewrite or migrate an implementation while preserving semantics, mechanically translate a large body of code, replace a runtime or language, run a strangler or big-bang port, turn compiler and test failures into a migration queue, or organize AI-assisted porting with one or many agents. Scale the workflow to the available time, compute, and model budget without weakening correctness gates. Do not use for ordinary dependency bumps, small local refactors, or product redesigns disguised as a port.
pr-review
Review and maintain GitHub pull requests end to end: inspect assigned or previously reviewed PRs, approve or request changes, handle author feedback, fix valid findings, recover CI, keep branches current, or classify caller-supplied review context without taking action. Use when a user asks to review PRs, catch up on PR work, handle review feedback, maintain their own PRs, names a PR number, requests a dry run, or delegates review-item analysis while retaining approval and delivery authority.
product-design
Plan, research, model, explore, prototype, and review digital product experiences. Use when turning product evidence or a provisional brief into a design brief; planning or synthesizing behavioral research to shape how the experience works; documenting a problem space, user expectations, or mental models; framing design opportunities; modeling objects, actions, relationships, journeys, and moments; creating information architecture, flows, wireframes, or prototype plans; or deciding how autonomy, learnability, fidelity, and visual character should shape an experience. Use for product and interaction design before or alongside implementation, especially when research, systems thinking, and interface craft must remain connected.
product-naming
Research, generate, screen, and shortlist names for products, software, services, companies, features, and internal initiatives. Use for naming or renaming work, multilingual name checks, spoken-name and spelling tests, domain and search-conflict research, preliminary trademark screening, naming scorecards, candidate shortlists, or a documented naming decision. Especially useful when a name must work across German and English or several European languages without sounding like a generic technology brand.
smart-dependency-updater
Research, select, introduce, group, implement, validate, and publish external dependency changes. Use when a user asks to add or choose a package, crate, action, image, SDK, or other versioned dependency; update or upgrade packages; replace Dependabot; make Renovate smarter; group dependency updates; create dependency PRs; assess changelog impact; modernize dependencies and related code; or adopt useful new APIs.
software-architecture
Assess, design, review, and evolve general software architecture with explicit system boundaries, responsibilities, data and integration contracts, quality attributes, operational concerns, and migration paths. Use when a user asks for software or system architecture, architecture options or review, service or module boundaries, monolith versus service decisions, scalability, reliability, deployability, operational readiness, performance objectives, load, soak, or stress objectives, workload scenarios, and capacity decisions, or non-frontend testing-strategy design such as the test pyramid, coverage goals, and which risks get which test types. Apply Twelve-Factor practices where they fit a long-running service; do not use for frontend-only architecture, codebase audit-only work, implementation plans, execution of an existing repository check, or recording an already-made decision when a narrower skill applies.
software-testing
Design, implement, diagnose, and verify focused non-frontend software tests for services, APIs, databases, async work, CLIs, and Rust. Use when asked to protect a behavior, invariant, regression, failure path, retry, authorization rule, migration, command-line contract, or flaky test; to diagnose test discovery, collection, runner, or framework configuration; to make a rule or state transition directly testable; to add a focused performance-regression guard; or to design, repair, or interpret a repository-native microbenchmark, comparative benchmark, or bounded end-to-end workflow benchmark. Prefer repository-native test conventions and real behavior over mock choreography. For browser, component, or React tests use effective-web. Do not use for visual, accessibility, or browser E2E testing; repository-wide coverage audits; testing-strategy design (use software-architecture); or load, soak, stress, and capacity methodology.
software-validation
Discover, execute, and report existing repository-native software checks. Use when asked to validate a change, run the repository's established checks against a change or package, run the applicable typecheck, static analysis, lint, format, test, benchmark, load, soak, stress, build, package, documentation, doctest, link, generated-reference, or example checks, or explain what validation evidence is still missing. Do not use to invent or install tooling, design new tests, choose architecture quality targets, prioritize repository improvements, or orchestrate delivery.
tech-docs
Design, write, update, and verify technical documentation for software users and contributors, including READMEs, setup and task guides, API and CLI references, migration notes, JSDoc, TSDoc, rustdoc, explanatory code comments, and executable examples. Use when asked to create or improve technical docs, document a public interface or migration, align documentation with implementation, or validate docs against repository conventions. Do not use for general marketing copy; route repository-wide documentation audits and prioritization to codebase-improvement.
Bio shown is the top-scored skill's repo description as a fallback — real GitHub bios land in a future update.