DobroslavRadosavljevic

Extract OpenGraph, Twitter Cards, and metadata from URLs or HTML, with dual extraction modes and social diagnostics for validator workflows.

Triggered when user asks to add structured logging, set up error tracking, implement monitoring, or configure observability. Automatically delegates to the logging-specialist agent.

Triggered when user asks to implement data validation, create validation schemas, add input validation, or ensure type-safe validation. Automatically delegates to the validation-specialist agent.

Triggered when user asks to design architecture, plan system structure, or make architectural decisions. Automatically delegates to the architect agent.

Triggered when user asks to audit code quality, assess maintainability, or review overall code quality. Automatically delegates to the quality-auditor agent.

Triggered when user asks to design APIs, create API specifications, or plan API endpoints. Automatically delegates to the api-designer agent.

Triggered when user asks to design database schemas, plan data models, or optimize database structure. Automatically delegates to the database-designer agent.

Triggered when user asks to design UI/UX, create user interfaces, or implement frontend designs. Automatically delegates to the ui-designer agent.

Build, review, or refactor React UI components into shadcn-style compound components. Use when creating composable component APIs, splitting prop-heavy components into named parts, wrapping Radix/Base UI primitives, adding asChild/Slot behavior, designing data-slot styling hooks, or enforcing accessible source-owned component primitives.

Implement-review-fix loop for development tasks. Use when the user invokes $loop, says "loop", "loop this", "run the loop", "implement then review", "review/fix until clean", or otherwise asks Codex to keep cycling through implementation, review, fixes, and re-review until no actionable issues remain.

Evidence-first research workflow for external web/source research and internal codebase investigation. Use when the user invokes $research, asks to research, investigate, audit what exists, compare options, understand current implementation, gather evidence, inspect docs, look up current facts, or answer repo-grounded questions before implementation. Prefer optional parallel agents for broad, separable research lanes such as external sources, codebase evidence, docs, risks, and competitor or alternative analysis.

Bans AI-slop code fingerprints and reworks implementation to match problem scale. Removes narrating comments, section banners, premature factories/Managers/Utils, try/catch wallpaper, silent fallbacks, any/as-unknown-as escapes, speculative “future” APIs, hallucinated imports, and unrequested platform features. Use when the user says "unslop-code", "AI code slop", "overengineered AI code", or asks to strip Copilot/Cursor/ChatGPT-looking code patterns.

Bans AI-slop public UI copy and rewrites website/app text into plain language for humans. Removes empty marketing verbs, robotic microcopy, vague CTAs, apology-padded errors, and especially leaked internals — agent instructions, implementation criteria, schema/API names, TODOs, stack traces, and prompt residue rendered as user-facing strings. Use when the user says "unslop-copywriting", "unslop copy", "AI copy", "robotic copy", "plain language UI", or asks to fix microcopy, marketing copy, errors, empty states, or text that reads like it was written for machines.

Bans AI-slop frontend aesthetics and reworks UI so it does not look machine-made. Removes purple/indigo gradients, Inter-class defaults, cream-terracotta-serif convergence, glassmorphism, pill badges, three equal feature cards, glow effects, emoji chrome, fake social proof, and template landing/dashboard skeletons. Use when the user says "unslop-design", "AI design slop", "looks AI", "generic AI UI", or asks to make frontend design distinctive, human-crafted, or free of AI-looking patterns.

Bans AI-slop documentation and rewrites README/docs into plain, verifiable guidance for humans. Removes emoji headers, badge walls, labeled-bullet Features, marketing fluff, install/API claims that don't match the repo, changelog empty phrases, unfit CONTRIBUTING boilerplate, and prompt/chat leakage. Use when the user says "unslop-docs", "AI README", "docs slop", or asks to fix documentation that looks generated, vague, or inaccurate.

Build, review, debug, migrate, or plan Legend-State v3 applications with current docs. Use for @legendapp/state v3 beta, observable, useValue, observer, Memo, $React, useObservable, computed or linked observables, observe, when, batch, fine-grained React rendering, React Native, local persistence, syncObservable, synced, syncedCrud, syncedFetch, syncedQuery, Supabase, Firebase, Keel, offline-first sync, transforms, syncState, testing, performance, and v2-to-v3 migrations.

Validate npm package names and check availability on the registry. Use when creating new npm packages, suggesting package names, or validating package naming conventions.