**3 issues found** across 2 files Confidence score: **3/5** - There is a concrete regression risk in `lib/admins/slack/fetchBotMentions.ts`: `topLevelMentionThreads` is written but never read, which suggests the deduplication flow is incomplete and could allow duplicate mention handling. - The `userCache` access pattern inside concurrent `Promise.allSettled` work in `lib/admins/slack/fetchBotMentions.ts` can trigger duplicate `getSlackUserInfo` calls for the same uncached user; this is likely non-fatal but adds correctness/performance uncertainty. - Given one high-severity, high-confidence logic issue plus a smaller concurrency concern, this looks like moderate merge risk rather than a hard block. - Pay close attention to `lib/admins/slack/fetchBotMentions.ts` - deduplication and cache behavior may cause duplicate mentions or redundant Slack user lookups.

Prompt for AI agents (unresolved issues)

```text Check if these issues are valid — if so, understand the root cause of each and fix them. If appropriate, use sub-agents to investigate and fix each issue separately. P1: Custom agent: **Flag AI Slop and Fabricated Changes** `topLevelMentionThreads` is populated but never read — it's dead code. This appears to be an incomplete deduplication mechanism: mentions that appear both at the top level and inside thread replies will be counted twice. Either implement the deduplication check (e.g., skip thread-reply mentions whose `channelId:ts` is in this set) or remove the variable. P2: Race condition on `userCache` within the concurrent `Promise.allSettled` batch. If two threads in the same batch contain replies from the same uncached user, both will call `getSlackUserInfo` before either populates the cache. Consider pre-populating the cache or serializing the user lookups to avoid redundant Slack API calls that count against rate limits. P3: `vi.resetAllMocks()` deviates from the project convention (`vi.clearAllMocks()`) documented in CLAUDE.md and used across all other test files. If stronger isolation is intentionally desired here, consider updating the project convention; otherwise, revert to `vi.clearAllMocks()` for consistency. ```

Architecture diagram

```mermaid sequenceDiagram participant Admin as Admin /content Page participant Service as fetchBotMentions participant Slack as Slack API participant Extractor as fetchThreadResponses Note over Admin,Extractor: Runtime flow for fetching bot usage metrics Admin->>Service: fetchBotMentions(period) loop For each channel Service->>Slack: conversations.history Slack-->>Service: List of messages Service->>Service: Filter top-level bot mentions Service->>Service: NEW: Identify messages with reply_count > 0 end Note over Service,Slack: NEW: Scan Thread Replies (Rate Limited) loop In batches of 5 (1.1s delay) Service->>Slack: NEW: conversations.replies (channel, thread_ts) Slack-->>Service: Thread message list Service->>Service: NEW: Find bot mentions inside thread opt Mentioning user not in cache Service->>Slack: users.info Slack-->>Service: User profile end end Note over Service,Extractor: Extract Bot Content (Videos/Tags) Service->>Extractor: CHANGED: fetchThreadResponses(token, targetTs) Note right of Service: targetTs is now parent thread_ts
for mentions inside threads loop For each mention Extractor->>Slack: conversations.replies (targetTs) Slack-->>Extractor: Bot responses in thread Extractor->>Extractor: Parse video URLs and tags end Extractor-->>Service: content data (links/tags) Service-->>Admin: Array of BotTag objects (sorted newest first) ```

_{Reply with feedback, questions, or to request a fix. Tag `@cubic-dev-ai` to re-run a review, or [fix all with cubic](https://www.cubic.dev/action/fix/pr/recoupable/api/409/ai_pr_review_1775570525323_5cb819ac-bb88-492f-ae8b-c4b4dde59197?entrySource=github_ui_to_cubic_ui).}