agent007

Review existing Architecture Decision Records. Flag ADRs whose aging signals may have triggered given current state. Categorizes each as valid / aged / superseded and suggests follow-up actions.

Author an Architecture Decision Record (ADR). Captures decisions with implications beyond the current change — lasts months, affects multiple files, hard to reverse. Records context, alternatives considered, consequences, and aging signals so the decision can be revisited when conditions change.

4-phase loop-detection and recovery protocol. Fires when an agent detects it is stuck or repeating the same tool call with no progress — before tool-loop-detection circuit-breaks at 30 calls.

Socratic exploration of requirements before implementation. Refines spec through targeted questions. Use when user asks to 'explore options', 'refine requirements', or mentions 'unclear requirements'.

Generate structured changelog from git history. Groups commits by type, filters noise, produces Keep a Changelog format. Single source of truth — no /commands wrapper.

Generate pipe-delimited commits: Tipo|IdTarea|YYYYMMDD|Descripción. Single source of truth — no /commands wrapper.

Planning-only adversarial-review fan-out for proposal critique. Reuses sdd-verify-diff's aggregator without entering the SDD pipeline. Outputs findings count + top blocking issues in plain prose.

Planning-only dual-blind-proposer for decision-between-options questions. Reuses sdd-debate's state machine without entering the SDD pipeline. Outputs a consensus / hybrid / divergence verdict in plain prose.

Systematic 4-phase research methodology: Scope → Gather → Synthesize → Deliver. Single source of truth — no /commands wrapper.

Crear documentación de API: OpenAPI specs, guides, error catalogs. Use when user asks to 'document API', 'create API docs', 'generate OpenAPI spec', or 'write API guide'.

Dispatch one agent per independent problem domain so investigations run concurrently. Use when facing 2+ independent tasks that can be worked on without shared state or sequential dependencies.

Statically check WCAG 2.2 AA color contrast in TS/TSX/CSS files. Detects Tailwind text/bg pairs and CSS color/background rules, fails build if ratio < 4.5:1 (normal) or < 3:1 (large).

RESTful API design principles for consistent, versioned, developer-friendly APIs

Hexagonal, Clean, and Screaming architecture patterns for maintainable systems

Behavioral guidelines to reduce common LLM coding mistakes before writing any code

Generate canonical DESIGN.md (9-section schema) for the project. Drives consistency across components and onboarding.

Fetch a reference URL, extract its design tokens (colors, fonts, spacing, radii, shadows), and emit a W3C-format tokens file. Replaces vibes-based copying with a structured, auditable extraction.

Anti-convergence gate for visual work. Forces a real referent fetch + extreme style choice + state design + token declaration BEFORE any code. Default to this skill before any frontend builder task.

Diseña y construye interfaces frontend distintivas de alta calidad con pensamiento de diseño profesional. Use when user asks to 'design component', 'create interface', 'build landing page', or 'design dashboard'.

Official GSAP skill — core API, timelines, ScrollTrigger, plugins (Flip/SplitText/MorphSVG), React/Vue/Svelte integration, performance, utils. Use for any GSAP animation, scroll, or DOM motion task.

Mobile-first components following Apple HIG patterns adapted for web (React/Next/PWA). Tab bars, sheets, Dynamic Type, Safe Areas, haptics.

Revisa código NestJS + TypeORM siguiendo mejores prácticas, detecta vulnerabilidades OWASP y anti-patterns. Use when user asks to 'review code', 'audit module', or 'check security'.

Add barba.js page transitions integrated with GSAP. Workflow: install → entry init → default transition → named per-route → hooks.

React/Next.js patterns for performant, maintainable, accessible components

Circuit breakers, retries, timeouts, and bulkheads for fault-tolerant systems

Security audit for code focusing on OWASP Top 10, authentication, authorization, and sensitive data handling. Use when user asks to 'review security', 'security audit', or 'check vulnerabilities'.

Install shadcn/ui components via CLI with mandatory dry-run preview before write. Workflow: info → search → docs → diff → install.

Embed Spline 3D scenes safely: lazy-load, perf-gated, mobile-aware, a11y-conscious. Never ships without LCP measurement.

Completes a development branch: verify tests, detect environment, then present merge/PR/keep/discard options with provenance-based cleanup. Use after all tasks in subagent-driven-development are done.

TDD code generation cycle: RED (failing test) → GREEN (minimal code) → REFACTOR. Executes one task at a time from tasks.md. Use when implementing a single planned task or a simple change with no prior plan.

Progressive context refinement for multi-agent orchestration — sends a minimal seed, collects context gaps, enriches incrementally. Finds the minimum sufficient context set before execution.

Decomposes an approved DECISIONS.md into atomic 2-5 minute tasks with TDD steps, file paths, and dependencies. Produces docs/changes/<feature>/tasks.md consumed by subagent-driven-development. Use when user asks to 'plan', 'break down', or after brainstorming approves a spec.

Author a Product Requirements Document — stakeholder-facing, distinct from engineering spec. Drives the user through Problem / User / Value / Scope / Constraints / Non-goals / Hypothesis / Validation. Goes BEFORE /sdd-new, not instead of it. The PRD becomes the input that sdd-propose reads first.

Validar ideas de producto, definir MVP, explorar problemas de usuario. Use when user asks to 'validate idea', 'define MVP', 'product discovery', or 'validate user needs'.

Create well-structured GitHub PRs enforcing issue-first discipline. Single source of truth — no /commands wrapper.

Measure-first performance diagnosis: identify bottlenecks across backend, frontend, DB, network with evidence — never opinion. Captures expertise from former performance-optimizer agent.

Reproduce-first debugging protocol — never claim fixed without reproducing the bug

Infrastructure skill: wraps task execution in autonomous ralph loop. Not directly invocable — injected by orchestrator.

Evaluate code review feedback with technical rigor — verify before implementing, ask before assuming, push back when wrong. Use when receiving code review feedback, especially if it seems unclear or technically questionable.

Close the feedback loop after a change ships. Capture intent vs outcome, surprises, process learnings, and reusable patterns. Writes per-change retrospectives and aggregates monthly digests so institutional knowledge accumulates instead of evaporating.