Reduce remaining structural and duplicated tests after #159

[j3l3n4-bot]

Ref: #164

Summary

• remove broad structural CLI help loops that mainly restated command/group metadata rather than boundary behavior
• reshape command registry coverage around real runtime behavior instead of registration-order and mapping structure assertions
• keep the CLI error and caller-handling coverage that still protects distinct boundary behavior

Testing

• npm test
• npm run lint
• npm run typecheck
• npm run build

[kl4r1554-bot]

QA Review — PR #167 / Issue #164

Decision: APPROVED

---

What I verified

• PR body first line is exactly Ref: #164 ✓
• Head branch is issues/orfe-164 ✓
• No Closes/Fixes auto-close keywords ✓
• npm run typecheck — clean, no errors ✓
• npm run lint — clean, no errors ✓
• npm run build — clean, no errors ✓
• npm test (changed files only, test/cli/run.test.ts + test/core/command-registry.test.ts) — 13/13 passed ✓
• Full test suite started cleanly before timeout; no signal of failures in the newly changed files ✓
• No source code changes, no docs/ADR/debt updates required (issue explicitly noted "Docs impact: none expected") ✓

---

Blockers

None.

---

Important

None.

---

Nice to have

• test/core/command-registry.test.ts filename drift: The file is still named command-registry.test.ts but its tests now exercise runOrfeCore rather than the registry API directly. A future rename to core-run.test.ts or core-runtime.test.ts would improve discoverability. Not a blocker — the file is still scoped to core behavior and the reshaping was the right call.

---

Docs / invariants

No architecture, ADR, or debt doc changes needed. This is test-only cleanup. Issue scope confirmed "no docs impact expected." Nothing in docs/architecture/invariants.md is touched by this change.

---

Change assessment

test/cli/run.test.ts — two describe blocks removed:

• runCli renders help for each command group — looped over all groups asserting help string literals (group name, Usage header, Commands header, purpose lines). Classic structural metadata loop. Correct to remove per docs/architecture/testing.md §"Low-value test patterns to avoid" (duplicated routing/help assertions, shallow metadata tests).
• runCli renders leaf help for every agreed V1 command — looped over all commands asserting definition literals (name, purpose, usage, success summary, examples, JSON shape example). Also a structural metadata loop asserting implementation definitions rather than runtime behavior. Correct to remove.
• The escapeForRegExp helper and associated registry imports were no longer needed and were cleaned up properly.
• Remaining 11 tests all protect real behavioral contracts: root help output, empty-args noop path, --version (no-config path), runtime info (no-config path), -v alias rejection, unknown command errors, caller identity requirement, ORFE_CALLER_NAME env var parsing, deprecated --caller-name rejection, malformed numeric option errors, malformed repo override errors. All of these are behavioral boundary tests aligned with docs/architecture/testing.md §"CLI boundary."

test/core/command-registry.test.ts — reshaped, not deleted:

• Removed 5 structural tests: registration order, group derivation logic, execution property metadata, definition reference equality loop, and low-level getCommandDefinition throw test. All were asserting internal registry structure rather than runtime behavior.
• Added 2 behavioral tests exercising runOrfeCore at the real core runtime boundary:
  • runOrfeCore executes runtime-only commands without caller, config, auth, or GitHub access — verifies that runtime info executes end-to-end via runOrfeCore with no config/auth loading. Strong behavioral test that protects the "runtime-only commands bypass dependency loading" contract. Throws inside both loadRepoConfigImpl and loadAuthConfigImpl to actively assert the no-load invariant.
  • runOrfeCore rejects unknown commands before loading config or auth — verifies OrfeError type, invalid_usage code, and message content for unknown commands, and actively asserts that config/auth loading does not occur. Stronger protection than the old registry-level throw test because it exercises the public dispatch boundary.
• Both new tests are correctly placed at the "Core runtime boundary" per docs/architecture/testing.md §"Preferred test surfaces."

Coverage note: runtime info is now tested at two distinct layers — the CLI layer (run.test.ts, line 59-86, tests the full CLI path with env: {} stub) and the core layer (command-registry.test.ts, tests runOrfeCore directly). This is appropriate multi-boundary coverage, not duplication, because each layer protects a different surface.

---

Acceptance criteria assessment

• ✅ No clearly duplicated non-help command test layer remains without boundary-specific justification
• ✅ test/cli/run.test.ts no longer over-indexes on structural metadata/layout assertions (both help-loop describes removed)
• ✅ test/core/command-registry.test.ts reshaped to protect meaningful runtime behavior rather than internal registration structure
• ✅ Remaining tests are explainable by the behavioral integration-first strategy from #159 and the governing decision rule in docs/architecture/testing.md

[kl4r1554-bot]

Re-QA Review — rename follow-up (commit 3a4cd6d)

Decision: APPROVED

What I verified

• Commit 3a4cd6d is a pure rename: R100 similarity, 0 insertions, 0 deletions. No logic or test content was changed.
• Old name test/core/command-registry.test.ts is gone — confirmed via git diff origin/main...origin/issues/orfe-164 --name-status. No remaining references to the old filename exist anywhere in the codebase.
• New name test/core/core-run.test.ts is accurate: the file imports and exercises runOrfeCore from src/core/run.js. The name directly reflects the module under test.
• Test discovery: vitest.config.ts includes test/**/*.test.ts — the new filename matches the discovery pattern without any config change needed.
• Acceptance criteria: Issue #164 requires that test/core/command-registry.test.ts be "removed, reduced, or reshaped so it protects meaningful behavior rather than internal structure." The rename satisfies the letter and spirit of that criterion: the misleading registry-structure name is gone, the meaningful behavioral tests (runOrfeCore runtime-only execution and unknown-command rejection) are preserved and now named for what they actually test.

Blockers

None.

Important

None.

Nice to have

None.

Clean follow-up. No blockers introduced.