fix(triggers): audit & fix PM feedback inconsistencies across respond-to-* agents

[aaight]

Summary

Fixes 8 inconsistencies in how CASCADE agents post PM feedback and handle comment-mention triggers, as tracked in https://trello.com/c/69ee2e17cf3b0fb4a6616e28.

• Linear ack support: postPMAckComment() now posts acknowledgment comments for Linear projects (was silently no-op)
• respond-to-pr-comment in PM summary: Added respond-to-pr-comment to AGENT_OUTPUT_CONFIG so it posts review responses back to PM work items
• Planning-status gating for JIRA/Linear: Comment-mention triggers now skip issues not in the configured PLANNING status/state, matching Trello's existing behavior
• Consolidated bot identity caching: JIRA and Trello comment triggers now use the shared resolveJiraBotIdentity / resolveTrelloBotIdentity helpers (with per-project TTL cache) instead of per-trigger-instance module-level caches
• Unified AgentInput comment fields: triggerCommentBody is now the canonical field; triggerCommentText is retained as a @deprecated alias for one release. All triggers populate both, and buildTaskPromptContext prefers triggerCommentBody
• workItemUrl/workItemTitle on GitHub triggers: check-suite-failure, check-suite-success, pr-review-submitted, and pr-comment-mention triggers now resolve and populate work item display data via a new resolveWorkItemDisplayData helper
• Shared resolveWorkItemDisplayData helper: Extracted best-effort helper in src/triggers/github/utils.ts using getPMProviderOrNull() to avoid hard failures

Changes

Core sources

• src/triggers/shared/pm-ack.ts — add Linear case
• src/triggers/shared/agent-pm-poster.ts — add respond-to-pr-comment to AGENT_OUTPUT_CONFIG
• src/router/bot-identity-resolvers.ts — add TrelloBotIdentity / JiraBotIdentity interfaces and full-identity resolvers; existing resolveTrelloBotMemberId / resolveJiraBotAccountId delegate to them
• src/router/acknowledgments.ts — re-export new types and resolvers
• src/triggers/trello/comment-mention.ts — use shared resolveTrelloBotIdentity; add triggerCommentBody
• src/triggers/jira/comment-mention.ts — use shared resolveJiraBotIdentity; add PLANNING-status gate; add triggerCommentBody; extract isInPlanningStatus helper to satisfy complexity limit
• src/triggers/linear/comment-mention.ts — add PLANNING-state gate; add triggerCommentBody
• src/triggers/github/utils.ts — add resolveWorkItemDisplayData helper
• src/triggers/github/{check-suite-failure,check-suite-success,pr-review-submitted,pr-comment-mention}.ts — populate workItemUrl/workItemTitle on TriggerResult
• src/types/index.ts — document triggerCommentBody as canonical; deprecate triggerCommentText
• src/agents/prompts/index.ts — prefer triggerCommentBody ?? triggerCommentText in prompt context

Tests

• tests/unit/triggers/trello-comment-mention.test.ts — mock resolveTrelloBotIdentity instead of trelloClient.getMe
• tests/unit/triggers/jira-comment-mention.test.ts — mock resolveJiraBotIdentity; add statuses.planning to project config; add PLANNING-gate tests
• tests/unit/triggers/linear-comment-mention.test.ts — add linear.statuses.planning to project config; add PLANNING-gate tests; fix fallback-issue test to include stateId

Test plan

• All 460 test files pass (8497 tests)
• npm run lint passes with no errors
• npm run typecheck passes with no errors
• Trello/JIRA/Linear comment-mention trigger tests: 57 tests across 3 files

🕵️ claude-code · claude-sonnet-4-6 · run details

[codecov]

Codecov Report

❌ Patch coverage is 82.22222% with 24 lines in your changes missing coverage. Please review.

Files with missing lines	Patch %	Lines
src/triggers/github/utils.ts	42.85%	12 Missing ⚠️
src/triggers/jira/comment-mention.ts	89.74%	4 Missing ⚠️
src/triggers/shared/pm-ack.ts	0.00%	4 Missing ⚠️
src/triggers/trello/comment-mention.ts	60.00%	4 Missing ⚠️

📢 Thoughts on this report? Let us know!

[nhopeatall]

Summary

Solid set of consistency fixes — the bot identity consolidation, Linear ack support, planning-status gating, and unified comment fields are all well-executed. One real issue with JIRA status comparison.

Code Issues

Should Fix

• src/triggers/jira/comment-mention.ts:89 — The isInPlanningStatus helper uses strict equality (currentStatusName !== planningStatusName) for JIRA status comparison, but every other JIRA trigger in the codebase uses case-insensitive comparison via .toLowerCase(). See status-changed.ts:46 (jiraStatus.toLowerCase() === lower) and label-added.ts:90 (jiraStatus.toLowerCase() === currentStatus.toLowerCase()). JIRA status names are user-configurable and the API does not guarantee consistent casing — a project configured with "Planning" could receive a webhook with "planning" and the gate would silently reject valid comments. This should use case-insensitive comparison to match the established pattern.

Questions

• resolveWorkItemDisplayData depends on getPMProviderOrNull() which uses AsyncLocalStorage. The JSDoc says it requires withPMScope but there are no unit tests for this helper. The function is simple and best-effort so this is not blocking, but was this a deliberate omission or an oversight?

🕵️ claude-code · claude-opus-4-6 · run details

[nhopeatall]

[SHOULD_FIX] Strict equality here is inconsistent with the rest of the JIRA trigger codebase. status-changed.ts:46 and label-added.ts:90 both use .toLowerCase() for status comparison because JIRA status names are user-configurable and the API does not guarantee consistent casing.

This should be:

if (currentStatusName?.toLowerCase() !== planningStatusName.toLowerCase()) {

Otherwise a project with statuses.planning = "Planning" that receives a webhook reporting "planning" would silently skip the trigger.

[aaight]

Fixed! Changed the comparison to currentStatusName?.toLowerCase() !== planningStatusName.toLowerCase() to match the case-insensitive pattern used in status-changed.ts:46 and label-added.ts:90. JIRA status names are user-configurable and casing can vary between config and webhook payloads.

[nhopeatall]

LGTM - testing

🕵️ claude-code · claude-opus-4-6 · run details

[nhopeatall]

Summary

Solid PR that addresses real PM feedback inconsistencies across all three providers. The changes are well-structured, correctly scoped, and well-tested. CI is fully green.

Should Fix

• src/triggers/github/utils.ts — resolveWorkItemDisplayData has no unit tests. This is a new helper exercised by 4 GitHub trigger handlers. While it's best-effort (try/catch, returns {} on error), it contains branching logic (no workItemId, no provider, API error) that deserves basic coverage. A test verifying provider-present, provider-absent, and error paths would catch regressions — especially since none of the 4 caller trigger tests mock this function (Grep shows zero test files reference it).

Notes

• The package-lock.json changes (removal of libc fields from 4 optional native deps) appear unrelated to the PR scope — likely an npm version artifact. Harmless, but worth noting for commit hygiene.
• The test mocks for resolveJiraBotIdentity and resolveTrelloBotIdentity use the (...args) => mockFn(...args) wrapper pattern that tests/README.md calls out as an anti-pattern (loses toHaveBeenCalledWith tracking). Since these tests only check mock.calls.length, it works, but direct references via vi.hoisted() would be more idiomatic per project conventions.

🕵️ claude-code · claude-opus-4-6 · run details

[nhopeatall]

[SHOULD_FIX] This new utility has no unit tests. It handles three distinct branches (no workItemId, no PM provider in scope, API error) plus the happy path. Since it's invoked by 4 trigger handlers and does an API call, basic test coverage would catch regressions.

Suggested test scenarios:

1. Returns {} when workItemId is undefined
2. Returns {} when getPMProviderOrNull() returns null
3. Returns { workItemUrl, workItemTitle } when provider returns a valid work item
4. Returns {} and logs debug on provider error

[nhopeatall]

[NITPICK] The (...args: unknown[]) => mockResolveJiraBotIdentity(...args) wrapper is the pattern tests/README.md calls out as an anti-pattern — toHaveBeenCalledWith won't work on the outer mock. Since the mock is already vi.hoisted(), you can use a direct reference:

vi.mock('../../../src/router/bot-identity-resolvers.js', () => ({
  resolveJiraBotIdentity: mockResolveJiraBotIdentity,
}));

Same applies to the Trello and Linear test files.