Fix/swarm memory resolve comment

[cuongdo-byterover]

Summary

• Problem:
• Why it matters:
• What changed:
• What did NOT change (scope boundary):

Type of change

• Bug fix
• New feature
• Refactor (no behavior change)
• Documentation
• Test
• Chore (build, dependencies, CI)

Scope (select all touched areas)

• TUI / REPL
• Agent / Tools
• LLM Providers
• Server / Daemon
• Shared (constants, types, transport events)
• CLI Commands (oclif)
• Hub / Connectors
• Cloud Sync
• CI/CD / Infra

Linked issues

• Closes #
• Related #

Root cause (bug fixes only, otherwise write N/A)

• Root cause:
• Why this was not caught earlier:

Test plan

• Coverage added:
  • Unit test
  • Integration test
  • Manual verification only
• Test file(s):
• Key scenario(s) covered:

User-visible changes

List user-visible changes (including defaults, config, or CLI output).
If none, write None.

Evidence

Attach at least one:

• Failing test/log before + passing after
• Trace/log snippets
• Screenshot/recording

Checklist

• Tests added or updated and passing (npm test)
• Lint passes (npm run lint)
• Type check passes (npm run typecheck)
• Build succeeds (npm run build)
• Commits follow Conventional Commits format
• Documentation updated (if applicable)
• No breaking changes (or clearly documented above)
• Branch is up to date with main

Risks and mitigations

List real risks for this PR. If none, write None.

• Risk:
  • Mitigation:

[github-actions]

Claude finished @cuongdo-byterover's task in 3m 10s —— View job

---

PR Review

• Gather context and read changed files
• Analyze types, adapters, coordinator, router, validator changes
• Review test coverage
• Post inline comments and summary

---

Summary

Score: 3.5 / 5

This PR resolves several legitimate issues from a prior review: lazy executor init in GBrainAdapter, infinite-loop protection in filename uniqueness logic, YAML injection in memory-wiki frontmatter, sync→async in the config validator, and as Error casts in the CLI commands. The direction is correct throughout. Three issues prevent a higher score.

---

What's good

• YAML injection fix (memory-wiki-adapter.ts:228): JSON.stringify(title) correctly handles ", newlines, and colons in titles that would previously corrupt the YAML frontmatter. This is the most important fix in the PR.
• Lazy executor init (gbrain-adapter.ts:164): The private get executor() pattern elegantly defers the synchronous execFileSync PATH probe to first-use. The added comment on line 60 documents the trade-off clearly.
• MAX_SUFFIX guards in both local-markdown-adapter.ts and memory-wiki-adapter.ts: The timestamp fallback is a safe unbounded-loop escape hatch.
• async validateGBrain in config-validator.ts: Removing execFileSync from the validator is correct — it was blocking the event loop during config startup.
• instanceof Error checks in curate.ts and query.ts: Replaces unsafe as Error casts with proper narrowing.

---

Issues

1. Import order in config-validator.ts (see inline comment, line 5)
const execFileAsync = promisify(execFile) is placed between import declarations. All imports should come first; the const should follow after.

2. Remaining as assertion in swarm-coordinator.ts (see inline comment, line 41)
(error as NodeJS.ErrnoException & {stderr?: string}).stderr uses as, which the project style guide prohibits (except lazy-init singletons). A simple {message: string; stderr?: string} cast or a 'stderr' in error guard would be idiomatic.

3. Missing tests for new behavior (see inline comments on swarm-coordinator.ts:29 and memory-wiki-adapter.ts:226)
Per the project's strict TDD policy, three new behaviors lack coverage:

• execBrvCurate should throw when content.length > 200_000.
• MemoryWikiAdapter.store() should produce valid YAML when the title contains ".
• MemoryWikiAdapter.store() filename collision fallback (suffix loop + timestamp).

These are correctness/security fixes — they should be verifiable by a test, not just by inspection.

---

Minor nits

• Redundant this.repoPath in gbrain-adapter.ts (inline, line 154): Now duplicates this.options.repoPath since options is stored as a field.
• Duplicate JSDoc in swarm-router.ts (inline, line 27): The old comment block for SELECTION_MATRIX was not removed when the new one was added.
• creative type removal: The routing matrix removal in swarm-router.ts is consistent and correct given the type narrowing in types.ts. classifyQuery never returned 'creative' anyway since there were no regex signals for it.

[danhdoan]

Lgtm