benji

Validates generated checklist covers all spec requirements. Ensures every FR has tests and implementation tasks.

You are a Claude Code hooks specialist, an expert in creating, configuring, and managing Claude Code hook configurations based on the official documentation. You specialize in hook configuration syntax, event types, matchers, security best practices, and workflow automation.

Validates test design quality at Phase X.0 (red phase) before implementation begins. Ensures tests are specific enough to guide correct implementation and catch real bugs. Use after writing tests, before starting implementation.

QC agent that validates Research skill outputs before commit. Checks interface source tracing, decomposition coverage, API documentation, and open questions.

Validates internal consistency of spec.md after parallel drafting. Checks that all sections reference the same FRs, types, and scope.

Orchestrates the spec and implement workflow for a single spec unit, maximizing parallelism via sub-agents. Dispatched by /orchestrate skill to execute a spec unit after research has created spec-research.md.

Validates implementation matches spec requirements. Verifies FR success criteria are met, interface contracts are followed, and scope is correct. Use at Phase X.2 Step 0 before code quality review.

QC agent that validates Spec skill outputs before commit. Delta check to ensure spec didn't break what Research QC validated.

Optimises tests for genuine confidence, not coverage metrics. Focuses on recently modified tests, ensuring they verify meaningful behavior and would fail if code were broken.

Spawns an Agent Team for divergent exploration of a problem space. Multiple teammates investigate in parallel, debate findings, and converge on a brief that feeds into /research. Use before /research when a problem needs multi-perspective discovery.

Implements specifications using TDD methodology with quality review. Use when implementing a spec, writing code from requirements, following TDD workflow. Includes test-first development, implementation, and thorough code review before PR.

Execution kickoff for feature work. Reads committed research artifacts, checks TaskList, and spawns spec-executor agents for parallel autonomous execution.

Touch up and refine UI/UX with live browser verification. Use for "polish this page", "touch up the UI", "peek at the styling".

Load essential project context and show current development status

Lightweight workflow for bugfixes and minor enhancements (<100 lines). Skips the full pipeline in favor of plan mode, focused implementation, and a single review pass. Use when the change is small and well-understood.

Explores a problem space, gathers context, designs approach, and decomposes into implementable specs. Use when starting new work, investigating issues, planning features, or decomposing large changes. This is the "fuzzy front end" where exploration and human dialog happen.

Deep PR review with inline GitHub comments. Launches parallel review agents, validates findings, drafts casual inline comments, and posts after user approval.

Creates a technical specification from research. Use after the research skill to produce spec.md and checklist.md for a decomposed spec unit.