Code quality agents, development utilities, and orchestration skills: architecture, security, QA, performance, test execution, code review, code simplification, plan adherence, LSP navigation, uv-python, incremental planning (with issue tracking and PR boundary markers), roadmap lifecycle management, session management, deep-research (with Bridged mode), business-panel, file-audit, bug-investigation, unfuck, swarm, quality-gate, pr-review, plan-review, map-reduce, speculative, reflect, fix, index-repo, summarize, and test-plan.
Prerequisite for /swarm and /unfuck: these skills spawn multiple named teammates that
communicate via SendMessage, which requires CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS=1 to be set
in the environment or settings.json.
Agents (8)
Agent
Purpose
Model
code-quality:architect
System architecture design and evaluation
Opus
code-quality:security
Application security review (OWASP)
Sonnet
code-quality:qa
Code quality, test strategy, tech debt
Sonnet
code-quality:performance
Performance profiling and optimization
Sonnet
code-quality:test-runner
Efficient test execution specialist
Haiku
code-quality:code-reviewer
Plan alignment, code quality, convention compliance, doc accuracy
Sonnet
code-quality:code-simplifier
Dead code removal, unnecessary abstraction cleanup, clarity improvement
Sonnet
code-quality:plan-adherence
Plan file verification, task completion tracking, file structure reconciliation
Opus
Skills (21)
Skill
Description
Type
/deep-research
Multi-hop research (40+ sources) with Bridged mode for internal+external analysis
Manual
/business-panel
Multi-stakeholder business analysis
Manual
/file-audit
Deep code quality audit system
Manual
/bug-investigation
Interactive bug hunting with background agents
PROACTIVE
/unfuck
Comprehensive one-shot repo cleanup
Manual
/swarm
Full agent team implementation
Manual
/quality-gate
Multi-pass review with adversarial lenses, fresh-context subagents, and blocking gates
PROACTIVE
/pr-review
Multi-agent PR review with finding verification
Manual
/plan-review
Multi-agent plan review with independent fresh-context reviewers
Manual
/map-reduce
Parallelized workload processing with chunking, mapper agents, and reducer synthesis
Manual
/speculative
Competing implementations in isolated worktrees with judge selection
Manual
/incremental-planning
Planning workflow with per-task review, assumption surfacing, and incremental PR boundary markers
Manual
/roadmap
Stateful multi-plan phase sequencing with roadmap lifecycle management: detect existing roadmaps, update, cleanup completed work, status/drift report, or fresh creation
Manual
/lsp-navigation
Semantic code navigation via LSP
PROACTIVE
/uv-python
Python tooling enforcement (uv over pip)
PROACTIVE
/test-runner
Efficient test execution patterns
Manual
/fix
Comprehensive finding fixer with spike execution for plan-review Research Gaps
Manual
/reflect
Mid-task self-reflection checkpoint via Serena metacognitive tools
Manual
/index-repo
Repository indexing for token-efficient codebase orientation
Manual
/summarize
Distills project-memory artifacts and PRs into human-readable summaries; optionally audits and archives
Manual
/test-plan
User-guided test plan generation with personas, Given/When/Then scenarios, manual UAT steps, and optional BDD .feature files
Manual
Commands (4)
Command
Description
/code-quality:session-start
Load project context or initialize new project
/code-quality:session-end
Sync project memory before ending
/code-quality:review-project
Comprehensive TODO validation
/code-quality:lsp-status
Check LSP server status
Requirements
At least one LSP plugin — Required for /lsp-navigation (pyright-uvx, vtsls-npx, gopls-go, etc.).
