rpikit

Use this agent to perform code review of implementation changes before security review. Reviews code for quality, design, correctness, and maintainability using Conventional Comments. Produces soft-gating verdict - REQUEST CHANGES allows proceeding with user approval.

Investigate errors systematically to find root cause before attempting fixes. Gathers evidence, analyzes patterns, and forms testable hypotheses.

Use this agent when you need to locate files in a codebase that are relevant to a specific task, feature, bug fix, or research objective. This agent excels at understanding the conceptual goal and mapping it to actual file locations, even when the user doesn't know exact file names or paths.

Use this agent to perform security review of implementation changes before completion. Reviews code modifications for vulnerabilities, insecure patterns, and security best practices. Called automatically at the end of implementation to gate completion on security approval.

Execute tests for a PR in isolated git worktree with comprehensive diagnostics

Run comprehensive verification checks (tests, lint, typecheck, build) before completion claims. Enforces evidence-before-claims discipline.

Use this agent when the user needs to conduct internet research on a specific topic, gather information from multiple sources, synthesize findings, or explore a research question. This includes fact-finding, competitive analysis, technology comparisons, learning about new concepts, or investigating specific questions that require web-based research.

Collaborative design methodology for creative work. Use before research or planning when requirements are unclear, multiple approaches exist, or the idea needs exploration. Refines ideas through progressive questioning.

Record architectural decisions as ADRs from design documents. Use after brainstorming or planning to capture what was decided, why, and what alternatives were considered. Produces sequentially numbered ADR files in docs/decisions/.

Structured completion workflow for implementation work. Use when implementation is complete, all tests pass, and you need to decide how to integrate the work. Guides merge, PR creation, or cleanup decisions.

Isolated workspace creation for parallel development work. Use when starting feature work that needs isolation from the current workspace. Creates git worktrees with proper setup and safety verification.

Disciplined plan execution with checkpoint validation, progress tracking, and verification at each step. Follows an approved plan strictly, running verification criteria before proceeding.

Concurrent agent dispatch for independent problems. Use when facing multiple independent tasks that can be worked on simultaneously. Reduces total time by parallelizing work that has no shared state.

Verification-first approach to code review feedback. Use when receiving review comments during the implement phase. Evaluate suggestions rigorously before implementing - no performative agreement.

End-to-end Research-Plan-Implement pipeline using parallel subagents. Each phase (research, plan, implement) runs in its own context window with file artifacts as the communication channel between phases.

Thorough codebase exploration that builds understanding through collaborative dialogue. Investigates architecture, patterns, and implementation details before planning or making changes.

Code review methodology for evaluating implementation changes. Use when reviewing code changes for quality, design, correctness, and maintainability. Focuses on changes made during implementation using Conventional Comments for clear, actionable feedback.

Security review methodology for evaluating implementation changes. Use when reviewing code changes for vulnerabilities, insecure patterns, and security best practices. Focuses on changes made during implementation rather than full codebase audits.

Consolidate multiple parallel research documents into a single unified report. Produces a self-contained document that a reader with no prior context can understand completely.

Root cause investigation methodology for bugs and failures. Use when encountering test failures, unexpected behavior, or errors during research or implement phases. Find the cause before attempting fixes.

Rigorous TDD methodology enforcing RED-GREEN-REFACTOR discipline. Use when implementing features or fixing bugs during the implement phase. Tests must be written before production code - no exceptions.

Evidence-before-claims discipline for implementation completion. Use before claiming any work is complete, fixed, or passing. Run verification commands and confirm output before making success claims.

Transform research findings into actionable implementation plans with granular steps, verification criteria, and stakes-based enforcement. Plans serve as contracts between human and AI.