name: e2e-testing description: Knowledge base for E2E test design and execution. Used by e2e-test-designer (catalog lookup), e2e-test-architect (new test design), and e2e-tester (execution) agents.

E2E Testing Skill

Purpose

Knowledge base for E2E test design and execution. Used by:

e2e-test-designer agent — Fast catalog lookup during planning
e2e-test-architect agent — Design new tests when no match exists
e2e-tester agent — Execute tests and report results

Agents That Use This Skill

Agent	Model	Purpose	When Invoked
e2e-tester	sonnet	Execute tests, report results	After milestone implementation
e2e-test-designer	haiku	Find existing tests, identify gaps	During /kdesign-impl-plan
e2e-test-architect	opus	Design new tests from scratch	When designer finds no match

How Tests Are Designed

Test design uses two agents with different cognitive demands:

/kdesign-impl-plan
    │
    ▼
e2e-test-designer (haiku) ─── catalog lookup
    │
    ├── match found ──→ return recommendation
    │
    └── no match ──→ e2e-test-architect (opus)
                          │
                          ▼
                    detailed test specification

e2e-test-designer (haiku) - Catalog Lookup

Receives validation requirements from /kdesign-impl-plan
Loads this skill's catalog
Searches for matching tests using heuristics
Returns recommendations OR hands off to architect

e2e-test-architect (opus) - New Test Design

Receives handoff from designer when no match exists
Analyzes requirements deeply
Designs comprehensive test specification
Returns detailed test with success criteria and sanity checks

Why two agents? Catalog lookup is mechanical (haiku is fast/cheap). Designing new tests requires reasoning about validation, false positives, and edge cases (opus provides quality).

See e2e-test-designer and e2e-test-architect for full details.

How Tests Are Executed

The e2e-tester agent:

Loads this skill (SKILL.md)
Finds requested tests in the catalog
Loads test recipe files
Runs pre-flight checks
Executes test steps
Reports PASS/FAIL with evidence

See e2e-tester agent for full details.

Test Catalog

Training Tests

Test	Category	Duration	Use When
training/smoke	Training	<30s	Any training changes
training/progress	Training	~60-90s	Progress tracking, metrics collection
training/cancellation	Training	~30s	Cancellation handling, cleanup
training/operations-list	Training	~5s	Operations API, list/filter
training/host-start	Training (Host)	~5s	Host service standalone
training/host-gpu	Training (Host)	~3s	GPU allocation
training/host-integration	Training (Integration)	~5s	Backend → host proxy
training/host-cache	Training (Integration)	~5s	Two-level operation ID mapping
training/host-completion	Training (Integration)	~5s	Full cycle through proxy
training/error-invalid-strategy	Training (Error)	~1s	Error handling
training/error-not-found	Training (Error)	~1s	Error handling

Data Tests

Test	Category	Duration	Use When
data/cache-get	Data	<3s	Cache loading, data retrieval
data/cache-range	Data	<100ms	Metadata queries
data/cache-validate	Data	<1s	Data validation
data/cache-info	Data	<500ms	Data inventory API
data/error-invalid-symbol	Data (Error)	<1s	Error handling
data/error-invalid-timeframe	Data (Error)	<1s	Error handling

Backtest Tests

Test	Category	Duration	Use When
backtest/smoke	Backtest	<10s	Any backtest changes
backtest/progress	Backtest	~20s	Progress tracking
backtest/cancellation	Backtest	~15s	Cancellation handling
backtest/api-start	Backtest (Integration)	~10s	API workflow
backtest/api-results	Backtest (Integration)	~15s	Progress via API
backtest/api-list	Backtest (Integration)	~10s	API cancellation
backtest/remote-start	Backtest (Remote)	~10s	Remote service
backtest/remote-proxy	Backtest (Remote)	~10s	Backend → remote proxy
backtest/remote-progress	Backtest (Remote)	~25s	Two-level progress
backtest/remote-cancel	Backtest (Remote)	~15s	Remote cancellation
backtest/error-invalid-strategy	Backtest (Error)	~2s	Error handling
backtest/error-missing-data	Backtest (Error)	~2s	Error handling
backtest/error-model-not-found	Backtest (Error)	~2s	Error handling

Agent Tests

Test	Category	Duration	Use When
agent/full-cycle	Agent	~2min	Full orchestrator cycle
agent/duplicate-trigger	Agent	~10s	Concurrency handling
agent/cancellation	Agent	~15s	Cancellation propagation
agent/status-api	Agent	~10s	Status API contract
agent/metadata	Agent	~2min	Metadata storage
agent/child-ops	Agent	~2min	Child operation tracking

CLI Tests

Test	Category	Duration	Use When
cli/client-migration	CLI (Migration)	~2min	Client consolidation M5 cleanup
cli/train-command	CLI (Restructure)	~60s	CLI restructure M1 - train command
cli/operations-workflow	CLI (Restructure)	~90s	CLI restructure M2 - operation commands
cli/information-commands	CLI (Restructure)	~30s	CLI restructure M3 - list/show/validate
cli/performance	CLI (Restructure)	~10s	CLI restructure M4 - startup performance

Pre-Flight Modules

Module	Checks	Used By
common	Docker, sandbox, API health	All tests
training	Strategy, data, workers	Training tests
backtest	Model, strategy, data, workers	Backtest tests
data	Directory access, mounts	Data tests

Reusable Patterns

(To be added in later milestones)

Troubleshooting

Domain	Module	Common Issues
Training	Training	Model collapse, 0 trades, NaN metrics, timeouts
Data	Data	Location issues, missing symbols, timeframe mismatch
Environment	Environment	Docker daemon, sandbox ports, resource exhaustion
Common	General	API schema, timeouts, permissions

Creating New Tests

Use TEMPLATE.md when creating new test recipes.

e2e-testing

SKILL.md