Skip to main content

Documentation Index

Fetch the complete documentation index at: https://mintlify.com/affaan-m/ECC/llms.txt

Use this file to discover all available pages before exploring further.

When the orchestrator agent (your main Claude session) needs to perform a deep security audit, resolve a Rust borrow-checker failure, or plan a multi-phase architecture decision, spawning a generalist is inefficient. ECC ships 67 pre-built agent definitions that encode narrow expert roles — each one scoped to a specific domain, toolset, and behaviour — so the orchestrator can delegate with confidence and recover clean, focused outputs.

What Agent Definitions Are

Agent definitions are .md files stored in agents/ and installed to ~/.claude/agents/. Each file encodes:
  • Role and identity — what the agent is and what it knows
  • Behaviour and scope — how it approaches tasks, what tools it may use
  • Output contract — what it produces and in what format
  • Allowed permissions — which tools, MCPs, and operations are in scope
The harness loads an agent definition when the orchestrator delegates a task that matches its role. The sub-agent runs in its own context with only the tools it needs, then returns a structured summary — keeping the orchestrator’s context window clean.
Sub-agents exist to save context by returning summaries instead of dumping raw tool output into the orchestrator. Pass objective context (the purpose behind a request, not just the literal query) when delegating, so the sub-agent can return answers that are actually useful to the orchestrator’s broader goal.

Agent Categories

Build Resolvers

Language-specific agents that diagnose and fix compiler errors, linker failures, and build tool issues. Each one knows the idioms, error messages, and resolution patterns for its target ecosystem.

Code Reviewers

18 language reviewers that apply language-specific linting, security, style, and correctness checks. Each reviewer knows the idiomatic patterns and anti-patterns for its language.

Architecture

Agents that reason about system design, service topology, and network layout. Used when decisions span multiple components or require deep structural thinking.

Security & Compliance

Agents that perform vulnerability analysis, policy compliance review, and healthcare-specific regulatory checks.

Ops & Orchestration

Agents that manage session flow, plan feature implementation, run end-to-end tests, and optimize the agent harness itself.

AI / ML

Agents specialized in GAN generation workflows, GAN training evaluation, and sub-agent output quality assessment.

Docs & Comms

Agents that keep documentation synchronized, look up API references, produce marketing content, and optimize SEO.

Open Source & Analysis

Agents for forking, packaging, sanitizing, and analyzing open-source projects.

Full Agent Roster

Build Resolvers

Invoked by /build-fix automatically when a build error is detected, or delegated directly by the orchestrator.
AgentTarget Ecosystem
build-error-resolverUniversal — detects language and routes to specific resolver
react-build-resolverReact / Next.js / Vite
rust-build-resolverRust / Cargo — including borrow checker and lifetime errors
go-build-resolverGo modules, go vet, build constraints
kotlin-build-resolverKotlin / Gradle / Android
java-build-resolverJava / Maven / Gradle
swift-build-resolverSwift / Xcode / Swift Package Manager
dart-build-resolverDart / Flutter
django-build-resolverDjango / Python dependency and migration errors
pytorch-build-resolverPyTorch / CUDA / dependency resolution
cpp-build-resolverC++ / CMake / linker errors
harmonyos-app-resolverHarmonyOS application build issues

Code Reviewers

Eighteen language-specific reviewers, each applying the idiomatic review criteria for its target language.
typescript-reviewer — Type safety, strict-mode compliance, async patterns, module boundaries.react-reviewer — Component patterns, hook rules, state management, performance.vue-reviewer — Vue 3 Composition API patterns, reactivity model, template best practices.
python-reviewer — PEP 8, type hints, idiomatic patterns, security, test coverage.django-reviewer — Django ORM patterns, view design, security middleware, migration review.fastapi-reviewer — Dependency injection, Pydantic validation, async patterns, OpenAPI compliance.
rust-reviewer — Ownership, lifetimes, unsafe usage, idiomatic error handling with Result/Option.go-reviewer — Idiomatic Go, concurrency safety, error handling patterns, interface design.cpp-reviewer — Memory safety, modern C++ idioms, RAII, concurrency, undefined behavior.
java-reviewer — Java design patterns, null safety, concurrency, Spring conventions.kotlin-reviewer — Null safety, coroutine safety, clean architecture, idiomatic Kotlin.
swift-reviewer — Swift concurrency, protocol-oriented design, memory management.flutter-reviewer — Flutter widget tree, state management, platform channel review.
php-reviewer — PHP security, modern PHP patterns, framework conventions.csharp-reviewer — C# idioms, async/await, LINQ, .NET conventions.fsharp-reviewer — Functional patterns, type-driven design, F# idioms.database-reviewer — SQL correctness, index usage, query performance, schema design.mle-reviewer — ML engineering review — model quality, pipeline hygiene, experiment tracking.

Architecture Agents

AgentRole
architectGeneral system design — service boundaries, data flow, scalability decisions
code-architectCode-level architecture — module design, dependency graphs, refactoring strategy
network-architectNetwork topology, routing, service mesh, infrastructure design
a11y-architectAccessibility architecture — WCAG compliance, semantic structure, ARIA patterns
homelab-architectHomelab / self-hosted infrastructure design

Security & Compliance

AgentRole
security-reviewerVulnerability analysis — OWASP top 10, injection, auth flaws, secrets detection
healthcare-reviewerHealthcare-specific compliance — HIPAA, data handling, PHI boundaries
network-config-reviewerNetwork configuration security review

Ops & Orchestration

AgentRole
chief-of-staffHigh-level session management — task routing, priority, orchestration decisions
plannerFeature implementation planning — breaks requirements into ordered, scoped steps
loop-operatorManages recurring agent loops and monitors loop health
harness-optimizerAudits and improves the agent harness configuration
e2e-runnerGenerates and runs Playwright end-to-end tests, captures screenshots and traces
tdd-guideEnforces test-driven development — scaffold interface → failing test → implement
refactor-cleanerRemoves dead code, consolidates duplicates, cleans up structure
performance-optimizerIdentifies and resolves performance bottlenecks
silent-failure-hunterHunts for silent failures — swallowed errors, missing error boundaries
pr-test-analyzerAnalyzes PR changes and generates targeted test cases

AI / ML Agents

AgentRole
agent-evaluatorEvaluates sub-agent outputs for quality and completeness
gan-evaluatorEvaluates GAN model outputs and training metrics
gan-generatorDrives GAN generation workflows
gan-plannerPlans GAN architecture and training strategy

Docs & Comms

AgentRole
doc-updaterKeeps documentation synchronized with code changes
docs-lookupLooks up API and library documentation via Context7
marketing-agentGenerates marketing copy, product descriptions, campaign content
seo-specialistSEO analysis, keyword optimization, structured data review

Analysis & Specialized

AgentRole
code-reviewerGeneral-purpose code quality, security, and maintainability review
code-explorerFast codebase navigation and structural analysis
code-simplifierIdentifies and simplifies unnecessarily complex code
comment-analyzerReviews code comments for accuracy and completeness
conversation-analyzerAnalyzes conversation patterns for insight extraction
type-design-analyzerReviews type system design and interface contracts
spec-minerExtracts specifications and acceptance criteria from requirements
network-troubleshooterDiagnoses network configuration and connectivity issues
opensource-forkerForks and prepares open-source projects for internal use
opensource-packagerPackages open-source projects for distribution
opensource-sanitizerSanitizes open-source projects — removes sensitive data, normalizes structure

Invoking an Agent Sub-Task

1

Automatic delegation via command

Commands like /build-fix and /orch-add-feature delegate to the appropriate agent automatically. You don’t need to name the agent — the command routes to the right one based on detected context (language, error type, task type).
2

Explicit delegation from the orchestrator

From your main session, instruct the orchestrator to delegate: “Use the security-reviewer agent to audit the authentication module.”The orchestrator will spawn the sub-agent with scoped context and return its findings as a summary.
3

Sequential phase orchestration

For complex workflows, chain agents through explicit phases:
Phase 1: RESEARCH  → code-explorer      → research-summary.md
Phase 2: PLAN      → planner            → plan.md
Phase 3: IMPLEMENT → tdd-guide          → code changes
Phase 4: REVIEW    → code-reviewer      → review-comments.md
Phase 5: VERIFY    → build-error-resolver (if needed) → done or loop back
Each agent gets one clear input and produces one clear output. Outputs become inputs for the next phase.
Scope your sub-agents tightly. Limited tools = focused execution and a smaller context footprint. Configure allowed tools and MCPs per agent for the best results.

Build docs developers (and LLMs) love