fix(triggers): pr-opened trigger respects reviewTrigger author modes

[zbigniewsobiecki]

Summary

• Gate PROpenedTrigger.matches() on ownPrsOnly/externalPrs — same pattern as CheckSuiteSuccessTrigger. When neither mode is active, the trigger won't match even if prOpened: true.
• Replace blanket isCascadeBot skip in handle() with author-based filtering — implementer-authored PRs fire only when ownPrsOnly is enabled; external PRs fire only when externalPrs is enabled.
• Add --pr-opened flag to review-trigger-set CLI — all review-adjacent triggers manageable from one command.
• Update dashboard trigger description to note that PR Opened respects Own PRs / External PRs modes.

Motivation

PR #522 (a human-authored dev → main merge) triggered respond-to-review even though reviewTrigger was configured as { ownPrsOnly: true, externalPrs: false }. Root cause: PROpenedTrigger was gated only by the prOpened boolean and ignored reviewTrigger entirely.

Test plan

• 20 unit tests (9 matches + 11 handle) covering all mode combinations
• Full suite: 183 files, 3049 tests passing
• npm run typecheck clean
• npm run lint clean

🤖 Generated with Claude Code

[aaight]

🤔 Just a sec, looking into that pr-opened trigger fix now

[aaight]

Thanks for taking a look! Let me know if you have any questions or feedback on the implementation.

The key changes are:

• PROpenedTrigger.matches() now gates on reviewTrigger config (same pattern as CheckSuiteSuccessTrigger)
• handle() filters PRs by author based on ownPrsOnly/externalPrs modes
• Added --pr-opened flag to the review-trigger-set CLI command
• 20 new unit tests covering all mode combinations

All tests passing and the implementation follows the existing patterns from the review trigger refactor.

[aaight]

🤔 Just a sec, looking into that PR review now

[nhopeatall]

Summary

LGTM — Clean bug fix that aligns PROpenedTrigger with the existing CheckSuiteSuccessTrigger pattern for respecting reviewTrigger author modes.

What I verified:

1. Pattern consistency: The matches() gate (!ownPrsOnly && !externalPrs → false) and handle() author filtering logic are structurally identical to CheckSuiteSuccessTrigger (lines 72–76 and 102–121 respectively). No new patterns introduced.

2. Legacy compatibility: The legacy fallback in resolveReviewTriggerConfig defaults ownPrsOnly: true when no reviewTrigger is present (via checkSuiteSuccess ?? true), so existing projects with { prOpened: true } and no explicit reviewTrigger continue to match. Covered by the "legacy config" test.

3. Behavioral change — personaIdentities undefined: Previously, missing personaIdentities caused the trigger to fire for all PRs (graceful degradation). Now it returns null. This is intentionally more conservative and matches CheckSuiteSuccessTrigger's behavior. Correct tradeoff — silent trigger fires when identity resolution fails were the root cause of the bug described in PR #522.

4. Reviewer persona handling: The old isCascadeBot check would skip PRs from both implementer AND reviewer personas. The new code only checks against the implementer login — reviewer-authored PRs are treated as "external" when externalPrs is enabled. This is correct: the reviewer persona authoring a PR is not the implementer, so the respond-to-review agent should fire on it. Test coverage exists for this case.

5. CLI change: --pr-opened correctly sends prOpened as a top-level key in the triggers payload (not nested under reviewTrigger), matching the GitHubTriggerConfigSchema structure. The updateTriggers mutation's z.record(z.union([z.boolean(), z.record(z.boolean())])) schema accepts this shape.

6. Tests: 20 tests all pass, covering all mode combinations including both-modes, neither-mode, legacy, and missing identities.