From claude-orchestration
Generate comprehensive tests for architectural layers with coverage-first analysis. Use when testing specific layers (core, domain, application, infrastructure, boundary). Reads testing strategy from playbook or uses interactive template selection.
How this skill is triggered — by the user, by Claude, or both
Slash command
/claude-orchestration:layer-testingThis skill is limited to the following tools:
The summary Claude sees in its skill listing — used to decide when to auto-load this skill
Orchestrate comprehensive layer testing with coverage analysis and isolated worktrees.
Orchestrate comprehensive layer testing with coverage analysis and isolated worktrees.
Coverage-first analysis prevents wasted effort—existing tests may already cover the code you're about to test. Running coverage before writing tests reveals exactly where gaps exist, letting you focus effort on uncovered paths rather than duplicating existing coverage.Isolated worktrees keep testing work separate from your working branch, preventing accidental commits to main development branches and enabling parallel testing of multiple layers.
- Write tests only—never modify production code during testing - Stay within the specified layer—do not test adjacent layers - Use worktree for all changes—keep working branch clean - Follow the testing strategy—do not add tests for explicitly skipped patternsFrom command invocation:
Testing Request:
- Module: auth
- Layer: infrastructure
- Playbook: docs/playbook.md (optional)
- Coverage Target: 100% (optional)
From natural invocation: Ask user for module, layer, playbook path (optional), coverage target (optional).
If playbook provided:
Read(${PLAYBOOK_PATH})
If no playbook: Ask user to choose:
<strategy_options>
If user picks 1-3: Use template from skills/layer-testing/templates/
If user picks 4: Ask:
Store strategy for agent use.
Check for existing tests:
find src/modules/${MODULE}/${LAYER} -name "*.test.*" -o -path "*/__tests__/*"
If NO tests: Skip to step 4
If tests exist:
Detect framework and run coverage (run these checks in parallel):
if [[ -f "vitest.config.ts" ]] || grep -q "vitest" package.json; then
pnpm test src/modules/${MODULE}/${LAYER} --coverage --reporter=json-summary
elif [[ -f "jest.config.js" ]] || grep -q "jest" package.json; then
pnpm test -- src/modules/${MODULE}/${LAYER} --coverage --coverageReporters=json-summary
fi
Parse coverage/coverage-summary.json:
TOTAL_COV=$(jq '.total.lines.pct' coverage/coverage-summary.json)
If 100% coverage:
✅ Already at 100% coverage!
Options:
1. Exit (nothing to do)
2. Verify test quality
3. Enhance tests
Choose (1-3):
If user picks 1, exit successfully.
If < 100%: Continue to Step 4
Find all files:
find src/modules/${MODULE}/${LAYER} -name "*.ts" ! -name "*.test.*" ! -path "*/__tests__/*"
Categorize using strategy:
Cross-reference with coverage (if available):
Show results:
📊 ${MODULE}/${LAYER}
Coverage: ${TOTAL_COV}% (target: ${TARGET}%)
Gap: ${GAP}%
✅ Fully Covered:
- File1.ts (100%)
⚠️ Partially Covered:
- File2.ts (45%, lines 10-20, 30-40 uncovered)
❌ Not Tested:
- File3.ts (0%)
⊘ Skip:
- schema.ts (per strategy)
Which files to test?
1. All gaps
2. Untested only
3. Partial only
4. Custom selection
Wait for user response.
BRANCH="test/${MODULE}-${LAYER}-coverage"
git worktree add ../worktree-${BRANCH} -b ${BRANCH}
Task({
subagent_type: 'general-purpose',
description: 'Test ${MODULE}/${LAYER}',
prompt: `
Test the ${LAYER} layer of ${MODULE} module.
Strategy: Read ${PLAYBOOK_PATH}
Files: ${SELECTED_FILES}
Target: ${COVERAGE_TARGET}%
Working directory: ${WORKTREE_PATH}
Follow strategy principles.
Write comprehensive tests.
Never modify production code.
Create single commit when done.
Report:
- Coverage achieved
- Tests created
- Any issues
`
})
Wait for agent completion.
# 1. Tests pass
pnpm test ${FILES}
# 2. Coverage meets target
pnpm test ${FILES} --coverage
# Parse and verify >= target
# 3. No production code changes
git diff --name-only | grep -v test | wc -l # Should be 0
# 4. Type check passes
pnpm typecheck
# 5. Lint passes
pnpm lint
All five gates must pass before proceeding to report.
✅ Testing Complete: ${MODULE}/${LAYER}
Coverage: ${BEFORE}% → ${AFTER}% (+${DELTA}%)
Tests: ${COUNT} new tests
Files: ${FILES_TESTED}
Commit: ${COMMIT_HASH}
Quality Gates:
✅ All tests passing
✅ Coverage target met
✅ No production changes
✅ Type check passing
✅ Lint passing
Next steps:
- Review tests in worktree
- Merge: git rebase ${BRANCH}
- Or: Continue testing other layers
npx claudepluginhub bengous/claude-code-plugins --plugin claude-orchestrationDesign test strategies and test plans with coverage targets. Complements /draft:coverage which measures what this skill plans. Auto-loaded by /draft:implement before TDD.
Creates and manages unit and integration tests by analyzing codebase, auto-detecting test frameworks, and generating tests that follow project conventions.