Plugin Claude Code
13 expert agents, 57 skills (flat depth-1), 43 deterministic hooks across 7 events — orchestration system with always-on behavioral contracts, SDD pipeline (proposal→spec→design→tasks→apply→verify→archive) with 4 auto-gates, V7.3 hardened auto-loop (fire→feedback→retry→converge/escalate), parallel fan-out with…
Instructions file
Claude Code instructions for SebastianDevps/agent007, covering agent007 v7 — orchestration system, core rules (always active), behavioral contracts (identity — always-on), routing (announce decision) and agent ↔ skill contract.
Agent
Agent "INDEX" from SebastianDevps/agent007, covering agents index — agent007 and removed in v6 refactor (2026-05-06).
Agent
Read-only macro-level architecture reviewer. Use PROACTIVELY before merging changes that cross module boundaries, alter public APIs, or introduce new bounded contexts. Distinct from code-reviewer (line-level) — focuses on boundary integrity, pattern consistency, cross-module concerns. Use PROACTIVELY when…
Agent
Senior backend & DB architect for NestJS/TypeORM/Postgres/Redis. Use PROACTIVELY for API design, schema decisions, migrations, performance, resilience. MUST BE USED for auth/payment writes. Use PROACTIVELY when: api, endpoint, database, query, sql, typeorm, microservice, cache, circuit-breaker, rate-limit.
Agent
Read-only senior code reviewer. MUST BE USED before merging any non-trivial diff — this is the quality gate. Filters ruthlessly — only ≥80% confidence findings are reported. Zero false positives. Use PROACTIVELY when: code review, check quality, review pr, quality check, audit code, review diff.
Agent
Synthesizes user-facing technical docs from SDD artifacts (archived changes, ADRs, PRDs, retrospectives). Use PROACTIVELY after /sdd-archive to publish README updates, getting-started guides, and changelog entries. Read-heavy, write-narrow. Use PROACTIVELY when: documentation, technical writing, onboarding doc, system…
Agent
Cross-subagent failure recovery commander. Use PROACTIVELY when a dispatched subagent returns BLOCKED, fails, or N consecutive subagents fail in the same wave. Classifies, recovers, escalates. Use PROACTIVELY when: subagent failure, cascading failure, recovery, BLOCKED status, multi-agent error, dispatch failure…
Agent
Senior frontend BUILDER & UX designer for React/Next.js/Tailwind. Use PROACTIVELY for component implementation, UI scaffolding, accessibility, performance. MUST BE USED for any visual work — defaults to BUILDER mode (writes code, not opinions). Use PROACTIVELY when: gsap, animation, scroll, shadcn, spline, barba…
Agent
Production incident commander for active outages, security breaches, and severity-1 events. MUST BE USED when a production incident is declared. Use PROACTIVELY when symptoms suggest customer-impacting failure. Use PROACTIVELY when: incident, outage, sev1, sev2, p0, p1, postmortem, downtime, breach, on-call, pager…
Agent
Autonomous ralph-loop control plane. Use PROACTIVELY when starting any iterative/until-X execution. Prioritizes safety over speed — a paused loop beats a runaway loop. Use PROACTIVELY when: loop, ralph, autonomous, until, iterate, run until, loop until, retry until, keep running, --persist.
Agent
Senior observability engineer for metrics, logs, traces, SLI/SLO design, and alert hygiene. Use PROACTIVELY when designing dashboards, defining SLOs, or diagnosing alert fatigue. MUST BE USED before shipping a new service to production. Use PROACTIVELY when: monitoring, prometheus, grafana, opentelemetry, slo-burn…
Agent
Senior DevOps & testing engineer for GitHub Actions, Docker, Kubernetes, Jest/Vitest, Playwright. Use PROACTIVELY for CI/CD design, deployment strategy, observability, test pyramid balance. Use PROACTIVELY when: deploy, docker, ci/cd, tdd, coverage, pipeline, kubernetes, infra, devops, playwright, jest, vitest.
Agent
Senior product manager for discovery, RICE prioritization, user stories, MVP scoping. Use PROACTIVELY before any new feature work to validate problem and acceptance criteria. Use PROACTIVELY when: product, roadmap, backlog, prioritize, rice, acceptance criteria, discovery.
Agent
Surgical dead-code detector and safe batch remover (knip/depcheck/ts-prune). Use PROACTIVELY before major refactors. Categorizes by removal risk, requires approval, verifies tests between batches. Use PROACTIVELY when: dead code, unused, clean up, remove unused, depcheck, knip, ts-prune, prune dependencies, cleanup…
Agent
Senior AppSec auditor for OWASP/JWT/PII/compliance. MUST BE USED for any auth, payment, encryption, or PII change. Use PROACTIVELY before merging code that touches sensitive paths. Use PROACTIVELY when: security, auth, jwt, oauth, owasp, vulnerability, permission, cors, xss, injection, compliance, secrets.
Command
Commands son orquestadores de alto nivel. Para skills invocables (commit, pull-request, changelog, deep-research, etc.) ver .claude/skills/INDEX.md.
Command
Route question to the correct expert agent based on keyword matching.
Command
Master development command — classifies task and routes to correct pipeline.
Command
Multi-agent workflow orchestration with named workflow types and structured HANDOFF documents between agents.
Command
Convierte intención vaga en un prompt-spec XML canónico (canonical XML conventions). Persiste en .sdlc/state/active-prompt.json para que subagent-context.py lo inyecte a cada delegación. Prioriza investigar → planificar → construir.
Command
Activates the Ralph Loop — autonomous iteration until the completion promise is detected. Can be invoked manually or auto-routed by the orchestrator.
Command
Local-only scan of the Agent007 ecosystem for configuration weaknesses and potential prompt injection vulnerabilities.
Command
CRUD sobre specs estables en openspec/specs/ — bypass del pipeline SDD para cambios chicos.