darachcawley Slack DM conversations by date
Produce a deduplicated list of DM-style conversations (1:1 IMs, self-DM, and MPDMs) where the authenticated Slack user had activity on a single calendar date.
Required input
- Date — Accept
YYYY-MM-DD, or relative terms (today,yesterday). Normalize toYYYY-MM-DDbefore calling Slack. If the user gives no date, ask once.
MCP server
Use the Slack integration configured in Cursor (commonly user-user-slack or user-slack). Substitute the real server name for {slack_mcp_server} in every tool call.
Before the first conversations_search_messages call: Read the tool descriptor JSON for that server (e.g. conversations_search_messages.json under the MCP’s tools/ folder) so limit, cursor, and parameter names match the schema.
Workflow
1 — Bulk search (minimal Slack calls)
Call conversations_search_messages on {slack_mcp_server}:
| Parameter | Value |
|---|---|
search_query |
is:dm |
filter_date_on |
Target date YYYY-MM-DD |
limit |
Maximum allowed by the tool schema (often 100) |
Paginate: If the response includes a cursor (or equivalent), repeat the same call with cursor until there is no next page or the response is empty. Optional safety cap (e.g. 50 pages); if reached, say so in the output.
2 — Dedupe by conversation
Treat each distinct value of the Channel column as one conversation (one row in the final list). Aggregate is enough—do not list every message unless the user asks for detail.
#mpdm-…— Group DM (MPDM).#U…(user id) — Typical form for a 1:1 DM channel in search exports.- If the MCP uses other
Channelshapes, still one row per distinctChannel.
3 — Labels (from search rows)
For each deduped Channel:
- 1:1: Prefer a title like
DM with @handleusing peerUserName/ RealName from any message row in that thread where the sender is not the requesting user (when identifiable). If only the user’s messages appear, useDM (peer unresolved)or resolve viausers_search/channels_list(im) only for that thread—do not fan out to every IM preemptively. - MPDM: Title like
MPDM: …using the#mpdm-…slug or, after a single optionalchannels_list(mpim, paginated), the Purpose / Topic line for a friendlier name. - Self-DM: If the conversation is the user’s notes-to-self channel, label
Self (notes)(match against workspace-rootuser_config.json→slack_usernameand/or the self IM fromchannels_listimnamed@{slack_username}if you already fetched lists for links).
4 — Optional archive links
If workspace-root user_config.json defines slack_dm_domain (and optionally slack_username for self-DM id resolution), you may add one Slack archive link per row after resolving D… or C… via channels_list (im / mpim, paginated). If config is missing or mapping is uncertain, omit links and still deliver the list.
Output format
Present to the user as a markdown table:
| Conversation | Kind | Notes |
|---|---|---|
| … | 1:1 / Self / MPDM | e.g. message count that day (optional) |
End with one line: date covered, count of distinct conversations, and whether pagination capped the search.
Optional config (workspace root)
user_config.json fields used when present:
slack_username— Self-DM labeling and exclusion from “with others” if the user later narrows scope.slack_dm_domain— Base host for optionalhttps://{slack_dm_domain}/archives/{id}/p…links.
Failure handling
- Tool / auth error: Report the error; do not invent conversations.
- Empty result after successful search: Report no DM/MPDM activity found for that date in search scope (valid outcome).
Relationship to daily digest
The daily-digest skill Step 3b instructs the agent to read this file and follow Workflow steps 1 through 4 below for DM/MPDM discovery on digest date D, while keeping all returned message rows grouped by Channel for downstream digest steps (thread replies, summarization, JSON)—not only the markdown inventory table in Output format.
When run standalone (user asks only for a DM list), follow Output format and you may aggregate per conversation. This skill does not replace Jira, configured channels, or viewer steps from daily-digest.