AI Test Generation at Scale

AI test generation with Stoneforge creates tests across your codebase in parallel. Instead of one developer writing tests file by file, multiple automated testing AI agents work simultaneously, each targeting different modules to boost your coverage fast.

Coverage audit, then parallel generation

Start by identifying the gaps. Stoneforge agents can scan your codebase, identify untested modules, and create a plan to cover them systematically.

# Create a test generation plan
sf plan create --title "Increase test coverage to 80%" \
  --description "Generate unit and integration tests for all modules below 50% coverage. Follow existing test patterns in __tests__/ directories."

# The Director creates scoped tasks:
# 1. Generate tests for src/services/auth/ (12% coverage)
# 2. Generate tests for src/services/billing/ (8% coverage)
# 3. Generate tests for src/api/routes/ (34% coverage)
# 4. Generate tests for src/utils/ (45% coverage)
# 5. Generate integration tests for API endpoints

Pattern-aware AI test generation

Agents study your existing tests before writing new ones. They pick up your testing conventions: which framework you use, how you structure test files, what mocking patterns you prefer. The goal is tests that look like your team wrote them.

# Configure workspace-level testing conventions
sf init

# In .stoneforge/prompts/worker.md, add:
# "Follow existing test patterns. Use Vitest with vi.mock().
#  Place test files adjacent to source files as *.test.ts.
#  Use factory functions for test data, not raw objects."

Every test runs before merge

Each generated test suite is executed in the agent’s worktree to verify everything passes. The Steward agent runs the full test suite again before merging, catching any interactions between newly generated tests — the same quality gate used in automated code review.

Target the riskiest code first

Focus AI test generation where it matters most. Prioritize modules that handle payments, authentication, or data integrity. Stoneforge’s priority system lets you ensure high-risk modules get tested first, reducing the chance of regressions in critical paths.

# High-priority: test the critical paths first
sf task create --title "Generate tests for payment processing" --priority 1
sf task create --title "Generate tests for auth middleware" --priority 1
sf task create --title "Generate tests for data export" --priority 2
sf task create --title "Generate tests for UI components" --priority 3