test(triggers): add unit tests for github trigger utilities and agent result handler

[zbigniewsobiecki]

Summary

This PR adds comprehensive unit tests for two critical utility modules in the trigger system:

• src/triggers/github/utils.ts — URL parsing and work item extraction logic (was at 53.22% coverage)
• src/triggers/shared/agent-result-handler.ts — Result artifact handling and cost accumulation (was at 0% coverage)

Test Coverage

github-utils.test.ts (27 tests):

• ✅ extractTrelloCardId — null input, no URL, valid URL with/without slug, multiple URLs, alphanumeric IDs
• ✅ hasTrelloCardUrl — truthy/falsy cases
• ✅ extractJiraIssueKey — null input, no key found, valid keys, embedded keys, word boundaries
• ✅ extractWorkItemId — Trello vs JIRA delegation, null handling
• ✅ requireWorkItemId — returns null with logging when no ID, returns ID when present

agent-result-handler.test.ts (9 tests):

• ✅ handleAgentResultArtifacts — cost accumulation, rounding to 4 decimals
• ✅ Skips when no cost field configured (Trello and JIRA)
• ✅ Skips when cost is zero or undefined
• ✅ Uses correct cost field for JIRA vs Trello projects
• ✅ Handles failed agent results with cost

Key Test Patterns

• Pure function testing for URL/text parsers (no mocks needed)
• Mocked getPMProvider and safeOperation for side-effect tests
• Follows existing tests/unit/triggers/ conventions
• All tests pass, lint clean, type checks clean

Related

Trello Card: https://trello.com/c/6993804d370da297fe1d6da8

[aaight]

🔍 Reviewing PR...

[aaight]

LGTM — Clean, well-structured unit tests that correctly exercise the source implementations.

I verified the tests against the source code in src/triggers/github/utils.ts and src/triggers/shared/agent-result-handler.ts:

• github-utils tests: Correctly test all exported pure functions against the actual regex patterns (TRELLO_CARD_URL_REGEX, JIRA_ISSUE_KEY_REGEX). Edge cases (null input, multiple matches, word boundaries) are well-covered. The extractWorkItemId and requireWorkItemId delegation tests correctly reflect the pm.type === 'jira' branching logic.
• agent-result-handler tests: The safeOperation mock correctly bypasses the wrapper to test the inner callback logic. The getPMProvider mock works correctly since it's called inside the safeOperation callback. The rounding test (1.123456 + 0.987654 = 2.1111) matches the actual Math.round(x * 10000) / 10000 logic. Both Trello and JIRA cost field resolution paths are tested, including skip-conditions (no field configured, zero cost, undefined cost).
• Test conventions follow existing patterns in tests/unit/triggers/.
• All CI checks pass.

One minor observation (not blocking): the comment on line 128 of github-utils.test.ts ("Should not match partial strings") is slightly misleading — the test actually demonstrates that word boundaries correctly truncate the match at NOTAKEY-123 rather than including -MORE. The expectation is correct; it's just the comment phrasing that's a bit ambiguous.