ingest-slack

ingest-slack — Slack-to-vault connector

Ingests recent Slack messages into the vault as markdown the graphify pipeline can read and the rest of the AI Brain Starter substrate (decision log, session-close cascade, hooks) can act on.

This is the first connector in a pattern. Adding the next external source (Notion, Jira, email) means writing a new normalizer, not a new architecture.

When to use

• User says /ingest-slack <channel-name> (with or without --days N)
• User asks to capture, sync, ingest, or pull a Slack channel into the vault
• User mentions wanting Slack threads available to the knowledge graph or session close

Do NOT use for:

• Sending Slack messages (use the Slack MCP slack_send_message directly)
• One-off thread reads (use slack_read_thread)
• Non-Slack sources (Notion, Jira, email get their own connectors)

How it works

1. Resolve the channel name to a channel ID via slack_search_channels
2. Fetch messages from the last N days (default 7) via slack_read_channel
3. For each parent message with replies, pull the full thread via slack_read_thread
4. Normalize messages to vault markdown (chronological, one ## per parent message, threads as ### sub-sections)
5. Write to External Inputs/Slack/<channel-name>/<YYYY-MM-DD>.md
6. Scan the new file for trigger keywords (exception, incident, pricing, escalation, outage, edge case, one-off, refund, refund request, custom deal, special pricing)
7. For each match, create a Decision Log stub at ⚙️ Meta/Decisions/<YYYY-MM-DD>-slack-<channel>-<sha8(message_ts)>.md
8. Print summary: file path, message count, edge cases detected

Voice rules

• No em dashes (use commas, colons, periods, parentheses)
• No exclamation marks
• Direct, no fluff
• Spanish messages stay in Spanish; English stays in English
• Author names quoted verbatim from Slack (display name, not real name)

Invocation

The skill is a thin orchestrator. The actual ingestion runs in Python at ~/.claude/skills/ingest-slack/ingest.py. The skill assembles the Slack MCP tool calls, hands the raw payloads to ingest.py, and the script does the normalization, file write, and edge-case scan.

When invoked:

1. Parse arguments: channel name (required), --days N (optional, default 7)
2. Call slack_search_channels with the channel name. If zero results, report "channel not found" and stop. If multiple, ask the user to disambiguate.
3. Call slack_read_channel with the resolved channel ID and a limit of 100. Compute oldest as the Unix timestamp for now - N days.
4. For each message in the response that has reply_count > 0, call slack_read_thread with the parent ts.
5. Hand the assembled payload (channel metadata + parent messages + threads) to ingest.py as JSON on stdin.
6. ingest.py writes the vault file, the decision stubs, and prints a summary.
7. Surface the summary to the user. If decision stubs were created, list the file paths so the user can review and fill in the outcomes.

Output contract

The vault file at External Inputs/Slack/<channel>/<YYYY-MM-DD>.md has frontmatter:

---
type: external-input
source: slack
channel: <channel-name>
channel_id: <Cxxxx>
date_range: <YYYY-MM-DD>..<YYYY-MM-DD>
message_count: <int>
ingested_at: <ISO 8601 timestamp>
---

Body is chronological. Each parent message is a ## YYYY-MM-DD HH:MM <author> section with the message body as its content. Thread replies become ### YYYY-MM-DD HH:MM <author> sub-sections.

Decision Log stubs created at ⚙️ Meta/Decisions/<date>-slack-<channel>-<sha8>.md carry frontmatter that matches the existing decision schema so the aggregator picks them up cleanly.

Idempotency

Re-running /ingest-slack <channel> --days N on the same calendar day overwrites the same vault file. The decision stub filenames hash the Slack message timestamp, so the same source message produces the same stub filename across re-runs. Re-runs do not duplicate; they refresh.

Acceptance test

A successful run produces:

1. One new (or refreshed) file at External Inputs/Slack/<channel>/<date>.md
2. Zero or more new decision stubs at ⚙️ Meta/Decisions/<date>-slack-<channel>-<sha8>.md
3. A stdout summary: Wrote N messages to <path>. Detected K edge cases. Stubs at: <paths>.

If the channel resolves but contains no messages in the date range, write the file anyway with message_count: 0 so re-runs are still idempotent and the absence is recorded.