Agent
Adversarial reviewer for technical documentation. Reports only what would cause a bad architecture, design, or implementation decision. Use on PRDs, ADRs, technical specs, design docs, RFCs, and system documentation before the work is implemented or relied upon. Spawned by the adversarial-doc-review skill — always in…
Command
Use when you need multiple alternative approaches before committing to a solution. Do NOT use when a clear solution exists (plan mode) or for comparing known options (/discuss:tradeoffs).
Command
Use to stress-test a proposal by hearing the strongest counter-arguments. Do NOT use for neutral analysis (/discuss:feature) or comparing options (/discuss:tradeoffs).
Command
Use before plan mode to critically analyze a feature proposal for gaps, edge cases, and risks. Do NOT use for simple Q&A, implementation, or code review (/code-review:).
Command
Use when you have 2-4 specific options and need a structured comparison. Do NOT use for open-ended exploration (/discuss:brainstorm) or single-option analysis (/discuss:feature).
Skill Claude CodeCodex
Use to adversarially review a technical document that already exists as a file — PRD, ADR, technical spec, design doc, RFC, or system/architecture documentation — to catch what would lead to a bad architecture, design, or implementation decision. Also use when asked to "review", "sanity check", "poke holes in", "red…
Plugin Claude Code
Feature development workflows: spec-driven development, test-driven development with coverage gates, and parallel codebase exploration.
Agent
Explores declared API contracts for a feature — OpenAPI, GraphQL, tRPC, protobuf, JSON Schema. Separate from endpoint code (backend-explorer covers that). Critical for multi-repo features. Opt-in primitive — invoke directly from any skill.
Agent
Explores backend/API layer for a feature. Finds models, views, serializers, endpoints, services, and database patterns. Spawned by explore-plan for parallel exploration.
Agent
Explores the configuration surface for a feature. Finds env vars, settings modules, feature flags, secrets handling, and per-environment overrides. Opt-in primitive — invoke directly from any skill.
Agent
Orchestrates execution of an approved PLAN-.md (or equivalent step list) by dispatching each step to plan-step-executor in order, accumulating carry-over notes between steps, and halting on blockers. Use AFTER a plan is approved and you want a single delegate to drive the whole feature instead of spawning…
Agent
Explores frontend/UI layer for a feature. Finds components, views, hooks, API clients, routing, and state management. Spawned by explore-plan for parallel exploration.
Agent
Explores git history and open PRs for a feature. Finds recent changes, conflicts, and historical patterns. Spawned by explore-plan for parallel exploration.