fix(github): match bot actors in comment filters using GraphQL __typename

[Neal006]

Fixes #1514

Problem

exclude_comments_by_actor and include_comments_by_actor never match any bot. This is broader than the issue title suggests: not only does the documented *[bot] wildcard fail, so does an exact entry like dependabot[bot]. Users configure bot filtering, see no error, and every bot comment still reaches the prompt.

Root cause

The two GitHub APIs disagree on bot logins:

API	login for the Dependabot app
REST	dependabot[bot]
GraphQL	dependabot (no suffix; identified by __typename: "Bot")

src/github/data/fetcher.ts filters comments from GraphQL data, but the patterns users write follow the REST/UI convention. So this check in actorMatchesPattern was dead code:

if (pattern === "*[bot]" && actor.endsWith("[bot]")) return true;

actor is dependabot, never dependabot[bot]. The suffix simply is not present in the data, which also breaks exact matches.

Why it was not caught

test/data-fetcher.test.ts mocked a payload GraphQL never produces:

author: { login: "scanner[bot]" }   // REST shape, not GraphQL

The test passed against fiction. It now mocks the real shape and fails without this fix.

Fix

1. Request __typename on the six Actor-typed author selections in src/github/api/queries/github.ts.
2. Add resolveActorName() to src/github/utils/actor-filter.ts, mapping an App actor to its suffixed name.
3. Use it at the single filtering choke point, filterCommentsByActor.

export function resolveActorName(
  author: { login: string; __typename?: string } | null | undefined,
): string {
  if (!author) return "ghost";
  if (author.__typename === "Bot" && !author.login.endsWith("[bot]")) {
    return `${author.login}[bot]`;
  }
  return author.login;
}

Normalizing at the filter boundary (rather than mutating fetched data) fixes the wildcard and exact-match cases in one place and leaves author names shown in the prompt unchanged. filterCommentsByActor is the single choke point for issue comments, PR comments, reviews, and inline review comments, so all four paths are fixed together.

Two details worth flagging for review:

• The commit author selection is deliberately left alone. It is a GitCommit (name/email), not an Actor, so __typename there would be meaningless.
• The existing ?? "ghost" behavior for deleted accounts is preserved inside resolveActorName, so null authors still never dereference.

Testing

Added 7 cases for resolveActorName in test/actor-filter.test.ts (bot suffixing, human passthrough, no double-suffix, ghost handling, absent __typename, plus wildcard and exact-match integration), and corrected the misleading mock.

Reverting only the src/ changes while keeping the corrected mock:

--- WITHOUT FIX ---
(fail) fetchGitHubData integration with time filtering >
       should filter reviews by both trigger time and actor
 93 pass, 1 fail

--- WITH FIX ---
 125 pass, 0 fail

Checks

• bun test test/data-fetcher.test.ts test/actor-filter.test.ts — 125 pass, 0 fail
• bun run typecheck — exit 0
• bun run format:check — clean
• Full suite: no regressions. Local run went from 849 pass / 44 fail to 858 pass / 42 fail; the remaining failures are a strict subset of the pre-existing ones on my platform (POSIX-only suites: downloadCommentImages, restoreConfigFromBase, SSH signing, path validation) and are unrelated to this change.

[Copilot]

Copilot was unable to review this pull request because the user who requested the review has reached their quota limit.