George-RD/claude-next-level

George-RD/claude-next-level

Agent Claude Code

Analyze blind comparison results to understand WHY the winner won and generate improvement suggestions.

George-RD/claude-next-level

Agent Claude Code

Compare two outputs WITHOUT knowing which skill produced them.

George-RD/claude-next-level

Agent Claude Code

Evaluate expectations against an execution transcript and outputs.

George-RD/claude-next-level

Agent

Use this agent for implementing features, fixes, or refactors in an isolated workspace. Spawned by the orchestrator after research is complete and the approach is decided. Always runs in an isolated workspace (JJ workspace or git worktree). Context: Research is done, user approved the approach for adding email digest…

George-RD/claude-next-level

Agent

Use this agent for exploring codebases, reading documentation, and gathering context before implementation or decision-making. Spawned by the orchestrator when a task requires understanding before action. Context: User files an issue about auth token refresh failing silently. user: "Investigate how token refresh works…

George-RD/claude-next-level

Agent

Use this agent for reviewing code changes on PRs or branches. Spawned by the orchestrator after an implementer finishes or when a PR needs review. Read-only — produces a review report but does not modify code. Context: Implementer just created a PR for the email digest feature. user: "Review PR #42 for code quality…

George-RD/claude-next-level

Agent

Between-task adversarial checkpoint during epic execution. Verifies completed task matches plan, checks for drift, validates integration.

George-RD/claude-next-level

Agent

Implementation agent for team execution. Implements a single task using strict TDD with quality hooks.

George-RD/claude-next-level

Agent

Adversarial review of implementation plans. Finds holes, missing edge cases, security issues, and over-engineering. Invoke during /spec-plan before user approval.

George-RD/claude-next-level

Agent

Adversarial review of project plans. Checks for missing requirements, unrealistic sequencing, scope creep, and integration risks. Invoke during /next-level:project before issue creation.

George-RD/claude-next-level

Agent

Quality and compliance review of spec implementations. Checks code matches the plan, tests are adequate, and code quality is good. Invoke during /spec-verify.

George-RD/claude-next-level

Agent

Use this agent for checking behavioral parity between source specs and target implementation during codebase porting. Spawned by the scheduler context during stage 5 (audit) to analyze modules in parallel. Context: Stage 5 — auditing ported code for behavioral parity. user: "Check parity for the auth module: spec at…

George-RD/claude-next-level

Agent

Note: For headless/loop execution, use claude -p --model haiku directly instead of this agent. This agent is for interactive use within a Claude session. Use this agent for extracting behavioral specifications from source code or test files during codebase porting. Spawned by the scheduler context during stages 1…

George-RD/claude-next-level

Agent

Use this agent for behavioral comparison between spec and implementation during retrospective audits. Spawned by the Phase 1 (codegap) prompt to process modules in parallel — one instance per module. Context: Phase 1 — analyzing behavioral gaps for the auth module. user: "Check gaps for auth module: spec at…