Skills

Designs and drives computer-use agents that operate a desktop or browser directly via screenshots and UI actions (click, type, scroll, navigate). Plans multi-step GUI workflows, handles verification and recovery between steps, and structures guardrails for safe autonomous execution. Built for models exposing computer use as a native client-side tool.

Finds, validates, and fixes security vulnerabilities end to end, going beyond audit-only review to produce verified patches. Reproduces the issue, proposes a minimal fix, and confirms the vulnerability is closed without breaking existing behavior. Models a CodeMender-style workflow where multiple agents detect, validate, and patch code security issues at scale.

Converts a design specification or screenshot into accessible, responsive production UI while preserving component boundaries, design tokens, semantic structure, and states.

Diagnoses failing browser and end-to-end tests by correlating assertions, traces, screenshots, network activity, timing, and application state into reproducible fixes.

Reviews cloud and AI infrastructure spend to identify measurable savings while preserving reliability, performance, security, ownership, and product requirements.

Builds regression suites for prompts and agent instructions so teams can detect quality, safety, format, and tool-selection regressions before deploying changes.

Converts a loosely defined opportunity into testable product requirements, user journeys, scope boundaries, assumptions, risks, success metrics, and discovery questions.

Crea evaluaciones repetibles para agentes de IA que miden el éxito de las tareas, el uso correcto de herramientas, la fundamentación, la latencia, el coste y la recuperación segura en escenarios realistas de extremo a extremo.

Convierte un repositorio desconocido en una guía de ingeniería concisa que explica la arquitectura, las convenciones, la configuración local, las rutas de ejecución principales y las primeras contribuciones de bajo riesgo.

Designs low-risk secret rotation for applications and agents, covering inventory, overlapping credentials, automation, audit evidence, emergency revocation, and rollback.

Optimizes prompts and agent workflows for finite context windows by prioritizing evidence, compressing history, managing retrieval budgets, and measuring token-cost trade-offs.

Simulates realistic adversarial behavior against AI features to uncover prompt injection, data exfiltration, unsafe tool use, jailbreak, and authorization failures before launch.

Produces clear, reviewable architecture decision records that capture context, alternatives, trade-offs, consequences, rollout steps, and reversal criteria.

Converts product safety requirements into practical agent and LLM policies with abuse cases, escalation paths, safe defaults, audit events, and acceptance criteria.

Audita interfaces de producto según requisitos prácticos de accesibilidad, incluidos la estructura semántica, el acceso mediante teclado, el comportamiento del foco, el contraste de color, los lectores de pantalla y los formularios.

Defines versioned data contracts between producers and consumers, with ownership, schemas, quality expectations, compatibility rules, and operational change management.

Turns merged work, pull requests, issues, and user-visible behavior changes into accurate, audience-specific release notes with breaking-change and upgrade guidance.

Tests AI agent tool use with realistic fixtures, malformed inputs, permission boundaries, deterministic assertions, and recovery checks for failed or partially completed actions.

Defines repeatable quality evaluation for LLM outputs using representative datasets, scoring rubrics, model-graded checks, human review sampling, and regression thresholds.

Plans safe database schema and data migrations with backwards compatibility, staged rollout, validation, rollback, performance safeguards, and application coordination.

Reviews Model Context Protocol servers for excessive permissions, unsafe tool design, prompt injection exposure, secret handling, authorization gaps, and auditability.