---
name: task-pipeline-sweep
type: task
description: Scheduled deployment of the Gmail→Pipeline sweep (Scout) — runs skill-pipeline-sweep once per intake (network, recruiters, roles, advisory). Gmail + Linear connectors; Pipeline team only. Orchestrator over skill-pipeline-sweep.
---

# task-pipeline-sweep

Scheduled deployment of the pipeline sweep. The engine is `skill-pipeline-sweep`; this task orchestrates the four intakes the engine must run, in order, each with its own source label, target funnel, and unit. This orchestration is the task's only added content — all extraction, enrichment, classification, dedup, filing, and reporting logic lives in the skill.

For each intake below, **read `.claude/skills/skill-pipeline-sweep/SKILL.md` and follow it verbatim** with the given parameters:

1. **Network** — query `label:pipeline/network -label:pipeline/triaged` → funnel **Network**, unit **person**.
2. **Recruiters** — query `label:pipeline/recruiter-internal -label:pipeline/triaged`, then `label:pipeline/recruiter-agency -label:pipeline/triaged` → funnel **Network**, unit **recruiter**.
3. **Roles** — query `label:pipeline/roles -label:pipeline/triaged` → funnel **Roles**, unit **role**.
4. **Advisory** — query `label:pipeline/advisory -label:pipeline/triaged` → funnel **Network**, unit **person**, default `opp:advisory`.

Run context: scheduled autonomous run on the Gmail + Linear connectors, Pipeline team only — never the delivery teams, a repo, or the Relay/exec/qa loop.

Produce **one combined report** across all four intakes: per-intake processed count (and remainder if capped), each new issue as `PIPE-NN — Title`, dedup comments, enrichment failures, bare-ID captures handed to Aled, classification uncertainties, and any wanted-but-missing labels omitted.

To change *how a sweep processes a thread*, edit `skill-pipeline-sweep`. Change the **set or order of intakes** here.