ompsure

Composure

Architecture & code quality

The foundational plugin. Composure enforces architecture standards, plans work before you build, and reviews changes with full structural awareness. It builds a code knowledge graph that powers instant discovery, blast radius detection, and test coverage prioritization across 8 languages. The task queue aggregates findings from all plugins, and the commit skill gates on unresolved items — nothing ships with known issues.

17 skills

Skills

/composure:initialize

Project Setup

One-time project bootstrap. Detects tech stack, queries Context7 for framework docs, builds code review graph, creates task queue infrastructure.

  • Generates .claude/no-bandaids.json config
  • Creates .code-review-graph/ database
  • Installs companion plugins (Sentinel, Testbench, Shipyard)
/composure:build-graph

Build Knowledge Graph

Creates or updates the code knowledge graph. Parses source files with tree-sitter, maps imports, identifies functions/types/components, builds dependency graph.

  • 8 languages: TypeScript, JavaScript, Python, Go, Rust, SQL, JSON, Markdown
  • 13 node types: File, Function, Class, Type, Test, Table, Column, RLS Policy, and more
  • Incremental updates via PostToolUse hook — always current
/composure:blueprint

Pre-Work Assessment

Graph-powered planning with progressive refinement. Classifies work, scans code graph for related files, confirms scope with user, analyzes impact, writes implementation spec.

  • 6 queries replace 76 agent calls (6.9x more efficient)
  • Progressive checkpoints shape the plan through conversation
  • Template forces per-file specs — catches design flaws during planning
/composure:review-pr

PR Review

Comprehensive PR review using the code graph. Identifies changes, resolves linked issues, builds structural context, generates walkthrough, runs quality audit.

  • Blast radius detection via graph edges
  • Cross-references Sentinel security findings
  • Verification discipline: comments with receipts
/composure:review

Delta Review

Token-efficient review of only changed code and its blast radius. Uses graph to auto-detect impact.

  • Structured risk assessment (low/medium/high)
  • Flags untested changed functions
  • Suggests verification approach
/composure:audit

Codebase Health Audit

Full audit across 5 categories: Architecture, Security, Code Quality, Dependencies, Test Coverage. Letter grades A-F.

  • Identifies files exceeding decomposition thresholds
  • Generates prioritized remediation plan
  • Outputs scored audit report with letter grades
/composure:code-organizer

Code Organizer

Restructures disorganized projects into conventional layouts. Detects file types, maps to appropriate directories, executes moves with import updates.

  • Analyzes 8 dimensions: directory inventory, misplaced files, mixed concerns, naming conventions
  • Framework-aware: knows where components, hooks, services, utilities belong
/composure:report

Suite Audit Report

Orchestrates all plugins to generate a self-contained HTML audit report with letter grades, tabbed sections, and prioritized recommendations.

  • Zero source code exposure
  • Professional visual report
  • Combines Composure + Sentinel + Testbench + Shipyard findings
/composure:commit

Commit with Gating

Commits with automatic task queue hygiene. Cleans resolved tasks, archives completed audits, blocks if staged files have open Critical/High items.

  • Auto-cleans [x] items from tasks.md
  • Archives fully-completed audit/blueprint files
  • Updates code graph post-commit
/composure:app-architecture

Architecture Router

Master router to framework-specific architecture guides. Detects stack, routes to frontend/fullstack/mobile/backend docs.

  • Phase 1-7 workflow: database → auth → queries → hooks → app shell → page → navigation
  • Integrates Context7-generated and team-written conventions
/composure:backlog

Task Processor

Processes accumulated code quality tasks. Modes: batch (sequential), delegate (parallel sub-agents), verify, archive, clean, sync.

  • Tasks sourced from PostToolUse hook, audits, blueprints, security scans
  • Severity prefixes: [DECOMPOSE], [SECURITY], [CVE-XXXX], [CI], [Docker]
/composure:changelog

Changelog Generator

Auto-generates changelog from git history enriched with code graph entity mapping. Groups by conventional commit type.

  • Maps changed files to business entities for context
  • feat, fix, refactor, test, docs grouping
/composure:view-graph

Graph Visualization

Opens interactive HTML visualization in browser. Files grouped by category, entity relationships, impact radius from selected nodes.

/composure:update-project

Config Refresh

Lightweight refresh for already-initialized projects. Updates version stamp, re-detects stack, regenerates stale docs.

/composure:integration-builder

Integration Builder

Stack-agnostic third-party integration scaffolding. Detects your project language, pulls latest SDK docs via Context7, and scaffolds client, auth, webhooks with proper error handling.

  • Supports TypeScript, Python, Go, Rust, and more
  • Tiered integration complexity (simple → enterprise)
  • Auto-generates test patterns for the integration
/composure:mcp-builder

MCP Server Builder

Build custom MCP servers to connect any service to Claude Code. Discovers the target API, designs tool definitions, and scaffolds a complete TypeScript MCP server project.

  • Discovers API via docs or OpenAPI spec
  • Generates typed tool definitions
  • Complete project scaffold with build + test
/composure:mcp-setup

MCP Setup

Discover and configure existing MCP servers for your project. Stack-aware recommendations from a curated registry with full setup — install, configure, and verify.

  • Curated registry of production MCP servers
  • Auto-detects which servers match your stack
  • Full setup: install, configure settings.json, verify connection

Measured Results

30x fewer tokens

Graph vs Explore Agents

5 targeted graph queries replaced 3 parallel Explore agents. ~500 words vs ~9,000 words. <5 seconds vs 45+ seconds.

2.75x efficiency

Blueprint vs Plan Mode

Graph-powered discovery + structured template. 74K tokens vs 204K tokens. 43 tool calls vs 133.

39/40 quality

Progressive Refinement

Template structure forces detailed specs that catch design flaws during planning. Surpassed Plan Mode quality.