test: Achieve 100% test coverage - all code must work

[EivMeyer]

Problem

Current test coverage is unacceptably low at 49.82% statements. Large parts of the codebase have 0% coverage:

Directory	Files	Current Coverage	Status
src/agent/	10	0%	UNTESTED
src/attach/	6	0%	UNTESTED
cli/	7	0%	UNTESTED
lib/	8	0%	UNTESTED
src/tui/	-	18.95%	Very Low
src/conductor/	-	60.36%	Medium
src/schemas/	-	67.56%	Medium

~13,500 lines of code are completely untested. This means bugs ship silently.

Why This Matters

• Untested code = unknown bugs - We have zero confidence agent execution, CLI commands, or attach protocol work correctly
• Refactoring is dangerous - Can't safely change code without tests proving it still works
• Current 1% threshold is meaningless - codecov.yml threshold provides zero enforcement

Implementation Plan

Phase 1: Agent System Tests (10 files, highest priority)

• agent-task-executor.test.js - Task execution flow, error handling
• agent-lifecycle.test.js - Start, stop, restart, cleanup
• agent-stuck-detector.test.js - Detection logic, thresholds
• agent-trigger-evaluator.test.js - Trigger conditions, edge cases
• agent-context-builder.test.js - Context assembly, validation
• agent-config.test.js - Config parsing, defaults, validation
• agent-hook-executor.test.js - Hook execution, failure handling
• output-extraction.test.js - Parse agent output formats
• output-reformatter.test.js - Output transformation
• schema-utils.test.js - Schema validation utilities

Phase 2: Attach System Tests (6 files)

• attach-client.test.js - Client connection, reconnection, errors
• attach-server.test.js - Server lifecycle, client handling
• socket-discovery.test.js - Socket discovery mechanisms
• protocol.test.js - Protocol message handling
• ring-buffer.test.js - Buffer operations, overflow handling
• index.test.js - Module exports, integration

Phase 3: CLI Tests (7 files)

• cli/index.test.js - CLI entry point, arg parsing
• commands/providers.test.js - Provider command handling
• message-formatters-normal.test.js - Normal output formatting
• message-formatters-watch.test.js - Watch mode formatting
• message-formatter-utils.test.js - Formatter utilities
• lib/first-run.test.js - First-run experience
• lib/update-checker.test.js - Update checking logic

Phase 4: Lib Tests (8 files)

• settings.test.js - Settings load/save/validate (416 lines!)
• provider-detection.test.js - Provider detection logic
• provider-names.test.js - Provider name resolution
• provider-defaults.test.js - Default configurations
• docker-config.test.js - Docker configuration handling
• id-detector.test.js - ID detection utilities
• stream-json-parser.test.js - JSON stream parsing

Phase 5: Coverage Threshold Enforcement

• Update codecov.yml threshold from 1% → 80% → 90% → 100%
• Add pre-commit hook that fails if coverage drops
• Add CI step that blocks merge if coverage < threshold

Testing Principles

1. Test behavior, not implementation - Verify outputs given inputs
2. Mock external dependencies - Docker, filesystem, network
3. Test error paths - Invalid inputs, timeouts, failures
4. Each test is independent - No shared state between tests

Acceptance Criteria

• All 31 untested files have corresponding test files
• Statement coverage ≥ 95%
• Branch coverage ≥ 90%
• Function coverage ≥ 95%
• CI blocks merge if coverage drops below threshold
• No /* istanbul ignore */ comments without justification

Related Issues

• refactor: Extract completion-detector logic into testable modules #53 - Extract completion-detector logic into testable modules
• Mock Docker operations in integration tests for faster local development #50 - Mock Docker operations in integration tests