Agent Selection Guide - Part 2: Specialized Agents
Contents
Use-Case Quick Reference for This Section
When to use this file:
- When choosing between search agents (hound, SERENA, context)
- When you need to select a code quality agent
- When you need to select a testing agent
- When you need DevOps or infrastructure agents
- When you need documentation agents
- When checking agent capabilities and tool expertise
Specialized Agents
Code Quality Agents
| Agent Name | Purpose | Input | Output | When to Use |
|---|
python-code-fixer | Fix Python lint/type/format issues | File paths | Issues fixed count | After editing Python files |
js-code-fixer | Fix JS/TS lint/type/format issues | File paths | Issues fixed count | After editing JS/TS files |
go-code-fixer | Fix Go lint/format issues | File paths | Issues fixed count | After editing Go files |
rust-code-fixer | Fix Rust lint/format issues | File paths | Issues fixed count | After editing Rust files |
Code Fixer Agent Rules:
- ALWAYS run after editing source files
- Run BEFORE committing code
- Can run in parallel (different files)
- Must report minimal output
Search and Analysis Agents
| Agent Name | Purpose | Input | Output | When to Use |
|---|
hound-agent | Deep file searches | Query + file patterns | Match count + file paths | Large files (>30KB), complex patterns |
log-auditor | Analyze log files | Log file path | Summary + error count | After test runs, CI failures, debugging |
context-agent | Find code elements from vague descriptions | Description | Element paths + snippets | Don't know exact names |
serena-agent | Find code elements by name | Symbol name | Element details + references | Know exact names |
Search Agent Selection Rules:
- Use
hound-agent for large files (>30KB): xml, html, json, txt, md, svg, css
- Use
log-auditor for log analysis after test runs
- Use
context-agent when you don't know element names
- Use
serena-agent when you know exact symbol names
Testing Agents
| Agent Name | Purpose | Input | Output | When to Use |
|---|
python-test-writer | Write Python tests | Module path + test count | Test count + file paths | Python test coverage needed |
js-test-writer | Write JS/TS tests | Module path + test count | Test count + file paths | JS/TS test coverage needed |
test-runner | Execute tests + audit logs | Test command | Pass/fail count + log path | Running test suites |
ci-monitor | Monitor CI/CD pipelines | Workflow name | Status + log path | Watching GitHub Actions |
Testing Agent Selection Rules:
- ALWAYS specify test count when using test-writer agents
- Default test count is 30 per function (too many!)
- Use
test-runner to avoid blocking orchestrator
- Use
ci-monitor for async CI/CD watching
DevOps and Infrastructure Agents
| Agent Name | Purpose | Input | Output | When to Use |
|---|
docker-agent | Docker operations | Dockerfile path + command | Container ID + status | Building/running containers |
kubernetes-agent | K8s operations | Manifest path + command | Resource status | Deploying to K8s |
terraform-agent | Infrastructure as Code | Terraform files + command | Apply status + resources | Cloud infrastructure |
github-actions-agent | GitHub Actions | Workflow path + event | Run ID + status | CI/CD automation |
DevOps Agent Selection Rules:
- Use specialized agents for each tool
- Never mix infrastructure tools in one agent
- Always validate manifests before apply
Documentation Agents
| Agent Name | Purpose | Input | Output | When to Use |
|---|
doc-generator | Generate docs from code | Source paths + format | Doc file paths | Auto-generating API docs |
readme-writer | Write README files | Project path + sections | README path | Creating project documentation |
tutorial-writer | Write tutorials | Topic + examples | Tutorial path | Creating user guides |
Documentation Agent Selection Rules:
- Use
doc-generator for API documentation
- Use
readme-writer for project READMEs
- Use
tutorial-writer for user-facing guides
Agent Capability Matrix
Core Capabilities
| Agent | Read Files | Edit Files | Run Tests | Fix Code | Git Ops | API Calls | Parallel Safe |
|---|
python-developer | Yes | Yes | Yes | Yes | Yes | Yes | Warning (git conflicts) |
python-code-fixer | Yes | Yes | No | Yes | No | No | Yes |
python-test-writer | Yes | Yes | Yes | No | No | No | Yes |
js-developer | Yes | Yes | Yes | Yes | Yes | Yes | Warning (git conflicts) |
js-code-fixer | Yes | Yes | No | Yes | No | No | Yes |
js-test-writer | Yes | Yes | Yes | No | No | No | Yes |
hound-agent | Yes | No | No | No | No | No | Yes |
log-auditor | Yes | No | No | No | No | No | Yes |
test-runner | Yes | No | Yes | No | No | No | Warning (test conflicts) |
Tool Expertise
| Agent | Linters | Formatters | Type Checkers | Test Frameworks | Build Tools |
|---|
python-developer | ruff, pylint | black, isort | mypy, pyright | pytest, unittest | uv, poetry, setuptools |
python-code-fixer | ruff, pylint | black, ruff format | mypy, pyright | - | - |
js-developer | eslint, tslint | prettier, dprint | tsc, flow | vitest, jest | pnpm, npm, webpack, vite |
js-code-fixer | eslint | prettier | tsc | - | - |
go-developer | golangci-lint | gofmt, goimports | go vet | go test | go build, make |
rust-developer | clippy | rustfmt | cargo check | cargo test | cargo build |
Domain Expertise (Star Rating: 1-5)
| Agent | Web Dev | CLI Tools | Data Science | DevOps | Mobile | Desktop |
|---|
python-developer | 3 | 5 | 3 | 4 | 1 | 2 |
js-developer | 5 | 3 | 2 | 3 | 2 | 3 |
go-developer | 3 | 5 | 1 | 5 | 1 | 2 |
rust-developer | 2 | 5 | 1 | 4 | 1 | 3 |
data-scientist | 1 | 2 | 5 | 1 | 1 | 1 |
ios-developer | 1 | 1 | 1 | 1 | 5 | 3 |
android-developer | 1 | 1 | 1 | 1 | 5 | 1 |
Related Files
- Part 1: Language Agents - Language-specific developer agents
- Part 3: Decision & Selection - Decision tree and selection checklist
- Part 4: Patterns & Practices - Anti-patterns and best practices
- Part 5: Advanced Topics - Advanced topics and troubleshooting
- Index - Main overview and quick reference