Zoom Meeting Context

Zoom Meeting Context

What this is, and isn't

NetClaw for Zoom (spec 118) has two separate recognition paths — don't confuse them:

1. Live in-meeting questions ("Toronto lost its BGP sessions, what happened?") are recognized automatically, inside zoom-rtms-mcp's own extractor.py, and routed straight to the Border's existing investigation path (bgp/federation/zoom_channel.py). This never invokes this skill.
2. This skill is for the historical-correlation half only (User Story 2): "didn't we see this before?" — searching the official Zoom Meetings MCP for a prior, real meeting and comparing what was discussed then against the network's actual current state.

MCP Servers

• zoom-rtms-mcp (NetClaw-authored, spec 118) — zoom_search_historical_meetings is a thin pass-through to the official Zoom Meetings MCP; use it, not a direct connection of your own.
• Zoom Meetings MCP (official, Zoom-hosted) — remote/OAuth, no local vendoring. Credential shape is still being confirmed against Zoom's own connector setup flow (tracked in specs/118-zoom-meeting-intelligence/research.md R6 / tasks.md T030).
• Whatever Member Claw already answers the "what's true right now" half (pyATS, NetBox, Splunk, etc.) — reuse the existing routing path, don't build a second one.

Workflow

1. Someone references a past meeting/incident by topic or approximate timeframe.
2. Call zoom_search_historical_meetings(query, time_hint).
3. If nothing relevant comes back, say so plainly. Do not present an unrelated result as if it were the match — this is an explicit spec requirement (FR-010's "no matching past meeting" case), not a style preference.
4. If a relevant past meeting is found, summarize what was discussed/decided then.
5. Check the network's current actual state for the same location/technology (via the existing Member-Claw routing — the same path zoom-rtms-mcp's live investigations use).
6. State plainly whether current state matches or differs from the historical discussion. Don't hedge past what the evidence actually shows.

What this skill must never do

• Never fabricate a historical match when the search comes back empty.
• Never treat this skill as a path to execute a configuration change — it is read-only correlation. Any actual change request, however it's phrased, still goes through NetClaw's existing device-write approval gate (Constitution Principles I–III), unchanged by this skill's existence.

Related

• specs/118-zoom-meeting-intelligence/ — full spec/plan/research/data-model/contracts
• mcp-servers/zoom-rtms-mcp/ — the live-listening half (extractor, panel feed, Border channel client)
• docs/ZOOM-MEETING-INTELLIGENCE.md — operator setup guide