Modules & Tools Dependency Map
How agentic modules build on each other across five layers, from orchestration workflows down to infrastructure tools.
Dependency map showing how agentic modules build on each other across five layers, from orchestration workflows down to infrastructure tools.
Layer 4 — Agents
Layer 3 — explore
Layer 2 — memory
Layer 1 — Leaf Utilities
Layer 0 — Tools
➔ Dependency arrow (module → depends on →)
Dependency Details
Each card shows a module and the other modules it explicitly references in its skill files, agent definitions, or tool implementations.
devteam
Multi-agent planning and implementation workflow. TeamLead orchestrates 8 subagents through structured plan→implement→review cycles.
Depends on: agent-communication explore · gather-context memory · search-memory format-md dev · documentation-rules
explore
Codebase exploration toolkit. The gather-context skill primes new sessions by pulling state from multiple analysis tools.
Depends on: git-report memory · search-memories graphify codebase-index tree
memory
Persistent knowledge vault with QMD-powered semantic search. Stores lessons learned, ADRs, PDRs, and auto-captures session state.
Depends on: qmd graphify
codebase-index
Semantic code search via Ollama embeddings. Indexes code chunks, supports incremental updates, and provides call-graph analysis.
Depends on: ollama
devbot (agent)
Pair programming partner agent. At session start, primes context via gather-context and searches memory for relevant past learnings.
Depends on: agent-communication explore · gather-context memory · search-memory memory · remember-session
docs
Architecture documentation generators. Auto-generates use-case maps from PHP codebases and provides an interactive application map editor.
Depends on: graphify python3 · ast-grep
Layer 4 — Agents
| Module | Description | Provides |
|---|---|---|
| devteam | Multi-agent team orchestration. Ships 9 agent profiles and 6 workflow skills (make-plan, review-plan, implement-plan, implement-story, review-implementation, summarize-plan). The central workflow engine. | 6 skills 9 agents |
| devbot (agent) | Pair programming partner agent. Works alongside human incrementally. Primes session context via gather-context and search-memory. | 3 agents |
Layer 3 — Explore
| Module | Description | Provides |
|---|---|---|
| explore | Codebase exploration toolkit. Three skills: gather-context, create-codebase-report, search-code. Depends on Layer 2 and Layer 1 utilities. | 3 skills |
Layer 2 — Memory
| Module | Description | Provides |
|---|---|---|
| memory | Persistent knowledge vault at .agents/memory/. Five skills: search-memory (QMD-powered), remember-session (session capture), memory-management (vault rules), thinking (scratch files), prune-memories (cleanup). | 5 skills 3 tools 7 hooks |
Layer 1 — Core Utilities
| Module | Description | Provides |
|---|---|---|
| agent-communication | Structured inter-agent protocol. Validates terminal status markers in messages. Used by devteam agents for reliable delegation handoffs. | 1 skill 1 tool hook |
| auto-recover | Detects transient provider errors and injects silent recovery prompts via session.error hooks. | 2 skills hook |
| format-json | Formats .json and .jsonc files via prettier with 2-space indentation. Auto-formats on file save. | 1 skill 2 tools hook |
| format-md | Formats markdown files via prettier. Aligns table columns, normalizes heading spacing. Auto-formats on save. | 1 skill 2 tools hook |
| format-yml | Formats .yml/.yaml files via prettier with 2-space indentation. Auto-formats on save. | 1 skill 2 tools hook |
| guards | Evaluates bash commands against configurable guard rules before execution. Prevents dangerous commands from running. | 1 skill hook |
| tree | Directory tree inspection tool. Shows full folder/file structure in markdown or JSON output. | 1 skill 1 tool |
| git | Git workflow skills: atomic commits, conventional commit messages, fixup commits, advanced history surgery, git-report. | 5 skills 1 tool |
| graphify | Codebase knowledge graph. Builds persistent graph with god nodes, community detection, and cross-file relationships. 4 CLI tools + MCP server. | 1 skill 4 tools MCP |
| websearch | Web search via Exa API. Retrieves current information beyond LLM knowledge cutoff. | MCP server |
| context7 | Official documentation retrieval for any library or framework. Fetches version-accurate code examples. | MCP server |
| architecture | Architecture governance. Five skills: architecture-rules, explicit-architecture, audit-codebase, audit-testsuite, make-adr. | 5 skills |
| dev | Development conventions hub — software-development (generic craft), address-review, rest-conventions, makefile, make-tests. | 5 skills |
| security | Security auditing with STRIDE threat modelling, OWASP Top 10 checklist, language-specific vulnerability checklists. | 1 skill |
| docs | Architecture documentation generators. Auto-generates use-case maps from PHP codebases using get-e/message-bus conventions. | 3 skills 2 tools |
| docker | Dockerfile authoring with battle-tested patterns for production-grade container builds. | 1 skill |
| k8s | Kubernetes manifest linting. Runs kubeconform (schema validation) and kube-linter (best practices). | 1 skill 1 tool 2 hooks |
| react | React 18+ and Next.js development conventions. Covers component design, server/client separation, TypeScript rules. | 1 skill MCP (next-devtools) |
| svelte | Svelte development conventions and best practices. | 1 skill MCP |
| chrome-devtools | Chrome DevTools MCP server. Inspect DOM, capture console errors, analyze network requests, run Lighthouse audits. | MCP server |
| playwright | Browser automation via Playwright MCP. Drive real browser sessions for end-to-end testing. | MCP server |
| signoz | SigNoz observability platform MCP integration. Dashboards, alerts, queries, investigation. Ships 2 commands: fix-n+1-problems, fix-slow-queries. | MCP server 2 commands |
| self-improvement | Meta-optimization module. 13 skills: create-skill, create-devbot-module, create-opencode-hook, create-opencode-tool, etc. | 13 skills |
| jetbrains | JetBrains IDE integration MCP. Code analysis, inspections, debugging, database tools, refactoring — all from within the IDE. | MCP server |
| tools-mcp | MCP server exposing devbot custom tools to agents. Each tool self-describes via mcp-meta subcommand returning JSON metadata. | 1 skill MCP |
| codebase-index | Semantic code search via Ollama embeddings. Find code by meaning, not keywords. Call-graph analysis. | 1 skill MCP |
| qmd | Semantic search over markdown vaults (BM25 + vector + hybrid + LLM reranking). | 1 skill 1 tool MCP |
| repomix | Directory packing into a single structured file for full-context analysis. | — |
Layer 0 — Tools (Infrastructure)
| Module | Description | Provides |
|---|---|---|
| ollama | Local LLM inference server running on port 18434. Powers embeddings for semantic code search and local model inference. | Docker |
| litellm | LLM proxy providing unified API access to multiple model providers. Disabled by default. | Docker |
| devbot (CLI) | CLI entry point — install, init, update, up, down, tool, module, models. | install init update |
| external-modules | Import external modules from github. Manage via devbot module add/remove/list/sync. | 1 tool |
External Modules
Third-party agentic modules registered via devbot module add. Live in vendor/, symlinked into .opencode/.
| Module | Source | Provides |
|---|---|---|
| addyosmani | github.com/addyosmani/agent-skills | 24 skills 4 agents |
| mindrally-svelte | github.com/mindrally/skills | skills |
| sveltekit-structure | github.com/spences10/svelte-skills-kit | skills |
| svelte5-best-practices | github.com/ejirocodes/agent-skills | skills |
| mindrally-react | github.com/mindrally/skills | skills |
| mindrally-nextjs | github.com/mindrally/skills | skills |
| mindrally-react-best-practices | github.com/mindrally/skills | skills |
| mattpocock-grilling | github.com/mattpocock/skills | skills |