shared/input-classifiers/slack-permalink

Source: shared/input-classifiers/slack-permalink.md

input classifier: Slack permalink

Pattern

1https?://<workspace>.slack.com/archives/<CHANNEL_ID>/p<TIMESTAMP>2https?://<workspace>.slack.com/archives/<CHANNEL_ID>/p<TIMESTAMP>?thread_ts=...

Where <TIMESTAMP> is the Slack ts field with the dot removed (e.g., p1715789432123456 → ts=1715789432.123456).

Fetch via

adk-mcp-slack (slack-mcp-server) — conversations.history + conversations.replies for the parent + thread.

If the MCP isn’t loaded (no SLACK_BOT_TOKEN and no SLACK_USER_TOKEN):

• Skip with [skipped] [slack] note in context.md.
• Do not fall back to scraping (no public Slack URL endpoint).

Extract into context.md

1### [slack] permalink — <channel-name> (<channel-id>)2url: <full URL>3fetched: <ts>4parent author: <user>5parent text (≤80 words): <quote with ≤15-word verbatim segments, paraphrase the rest>6thread length: <N replies>7thread excerpt: <last 3 substantive, quoted ≤15 words each>8references in thread: <PR URLs, Jira keys, DD links found in the thread>

Hints

• /adk-investigate: alert permalinks are common starting points. Extract symptom + symptom-time.
• /adk-implement: thread may contain refinement of a spec.
• /adk-document --type slack-summary: condense a long thread.

Privacy

• Don’t include user emails or message bodies verbatim in long quotes. Paraphrase.
• DM channels (D...-prefixed channel IDs): only fetch if the user explicitly says so.