ktx/docs/superpowers/plans/2026-05-15-claude-code-agent-runner-cli-test-closure.md

Claude Code Agent Runner CLI Test Closure Implementation Plan

For agentic workers: REQUIRED SUB-SKILL: Use superpowers:subagent-driven-development (recommended) or superpowers:executing-plans to implement this plan task-by-task. Steps use checkbox (- [ ]) syntax for tracking.

Goal: Restore the CLI ingest verification suite after the Claude Code agent-runner backend changed the no-LLM guidance text.

Architecture: Keep the implemented runtime behavior unchanged. Update the stale CLI ingest test expectation so it matches the current context-runtime guard that accepts either a configured global LLM provider, llm.agentRunner.backend: claude-code, or an injected agent runner.

Tech Stack: TypeScript, Vitest, pnpm.

---

Implemented spec coverage

The previous Claude Code backend plans are implemented in source: llm.agentRunner.backend exists, the final agent tool map uses AgentToolSet, the Claude Agent SDK runner disables built-in tools, filesystem settings, and SDK skills, ingest and memory DI choose the Claude runner, setup probes Claude Code auth with accountInfo(), and WorkUnit SDK tool failures enter the existing transcript-based failure counter.

The remaining v1 blocker is verification-only: packages/cli/src/ingest.test.ts still expects the pre-agent-runner error text, so pnpm --filter @ktx/cli exec vitest run src/ingest.test.ts fails even though runtime behavior is correct.

File structure

• Modify packages/cli/src/ingest.test.ts to expect the Claude Code-aware setup guidance.

---

Task 1: Update the stale CLI ingest assertion

Files:

• Modify: packages/cli/src/ingest.test.ts

• Step 1: Confirm the existing failure

Run:

pnpm --filter @ktx/cli exec vitest run src/ingest.test.ts

Expected: FAIL in prints provider setup guidance when a skip-llm setup project runs ingest because the test expects:

ktx ingest requires llm.provider.backend: anthropic, vertex, or gateway, or an injected agentRunner.

but the runtime now emits:

ktx ingest requires llm.provider.backend: anthropic, vertex, or gateway; llm.agentRunner.backend: claude-code; or an injected agentRunner.

• Step 2: Update the expected guidance string

In packages/cli/src/ingest.test.ts, replace the stale assertion:

    expect(runIo.stderr()).toContain(
      'ktx ingest requires llm.provider.backend: anthropic, vertex, or gateway, or an injected agentRunner.',
    );

with:

    expect(runIo.stderr()).toContain(
      'ktx ingest requires llm.provider.backend: anthropic, vertex, or gateway; llm.agentRunner.backend: claude-code; or an injected agentRunner.',
    );

Keep the following assertion that checks the Anthropic setup command:

    expect(runIo.stderr()).toContain(
      `ktx setup --project-dir ${projectDir} --anthropic-api-key-env ANTHROPIC_API_KEY --anthropic-model claude-sonnet-4-6 --no-input`,
    );

• Step 3: Run the focused CLI ingest test

Run:

pnpm --filter @ktx/cli exec vitest run src/ingest.test.ts

Expected: PASS with Test Files 1 passed and Tests 30 passed.

• Step 4: Run the Claude Code setup regression tests

Run:

pnpm --filter @ktx/cli exec vitest run src/setup-models.test.ts src/index.test.ts

Expected: PASS with Test Files 2 passed and all tests passing.

• Step 5: Run CLI type-check

Run:

pnpm --filter @ktx/cli run type-check

Expected: PASS with no TypeScript diagnostics.

• Step 6: Commit

Run:

git add packages/cli/src/ingest.test.ts
git commit -m "test: update claude-code ingest guidance expectation"

Expected: Commit succeeds with only packages/cli/src/ingest.test.ts staged.

---

Self-review

• Spec coverage: This plan closes the only v1-blocking gap found in the audit, a failing CLI test caused by stale expected text after the implemented Claude Code agent-runner guard changed.
• Placeholder scan: The plan contains exact file paths, commands, expected outputs, and replacement code.
• Type consistency: The expected string matches the implemented guard in packages/context/src/ingest/local-bundle-runtime.ts.