No E2E integration tests — safeguards validated in isolation only

[Shatla-tech]

Problem

The test suite has 276 unit tests covering individual components (circuit breaker state transitions, rate limit counters, response parsing) but zero end-to-end integration tests that validate safeguards working together with real Claude Code execution.

From TESTING.md, E2E tests are listed as "0 of 10+ target".

Impact

Safeguards could individually pass all unit tests while failing in production scenarios:

• Circuit breaker may not trigger correctly when Claude produces realistic output
• Rate limiter + circuit breaker interaction untested
• Exit conditions not validated against real Claude response formats
• Permission denial handling not tested with actual Claude CLI

Suggested Fix

Add a mock Claude CLI mode for CI testing:

```bash

In ralph_loop.sh, support RALPH_MOCK_CLAUDE for testing

if [[ "${RALPH_MOCK_CLAUDE:-false}" == "true" ]]; then
CLAUDE_CODE_CMD="$RALPH_DIR/../tests/mock_claude.sh"
fi
```

`tests/mock_claude.sh` can simulate realistic scenarios:

• Normal completion (outputs EXIT_SIGNAL: true after N loops)
• Stuck loop (never outputs progress)
• Permission denial
• High token usage response
• False-positive completion keywords in documentation

This allows full E2E tests in CI without requiring a real Claude API key or incurring costs.

[coderabbitai]

🔗 Related PRs

#52 - test(cli): add comprehensive CLI argument parsing tests [merged]
#60 - test(setup): add 36 comprehensive tests for setup.sh [merged]
#142 - feat(exit): detect permission denials and halt loop (Issue #101) [merged]
#191 - feat(loop): validate Claude Code CLI command at startup (#97) [merged]
#215 - fix(loop): detect is_error:true in JSON output despite exit code 0 [merged]

---

📝 Issue Planner

_{Check the box below or use the @coderabbitai plan command to generate an implementation plan and prompts that you can use with your favorite coding assistant.}

• Create Plan

---

🧪 Issue enrichment is currently in open beta.

You can configure auto-planning by selecting labels in the issue_enrichment configuration.

To disable automatic issue enrichment, add the following to your .coderabbit.yaml:

issue_enrichment:
  auto_enrich:
    enabled: false

💬 Have feedback or questions? Drop into our discord!

[github-actions]

Triage: Enhancement / Feature Request ✨

This issue has been categorized as an enhancement — it proposes adding E2E integration tests using a mock Claude CLI mode, which is a new capability rather than a bug fix.

Assessment

The issue is well-scoped and actionable:

• Clearly identifies a testing gap (0 E2E tests vs. 10+ target per TESTING.md)
• Provides a concrete implementation approach (RALPH_MOCK_CLAUDE env var + mock_claude.sh)
• Lists specific scenarios to cover (normal completion, stuck loop, permission denial, etc.)
• Aligns with the E2E testing philosophy already documented in CLAUDE.md (Playwright for UI, real services where possible)

Suggested Next Steps

1. Implement tests/mock_claude.sh with scriptable output scenarios
2. Add RALPH_MOCK_CLAUDE support in ralph_loop.sh (guarded so it never activates in production)
3. Write E2E bats test files under tests/integration/ or tests/e2e/ exercising the scenarios listed
4. Update TESTING.md to reflect the new E2E test count once implemented

This would be a great contribution — the TESTING.md already acknowledges E2E coverage as a known gap.

Generated by Issue Triage Assistant for issue #225

[frankbria]

Related: #152 (integration tests for task import functions). Both address test coverage gaps — #152 is narrower (unit/integration tests for lib/task_sources.sh), this issue is broader (E2E loop tests with real Claude Code execution). Suggest implementing #152 first as it's more self-contained, then using those patterns when building out the E2E suite here.