OSS maintainer fleet

Maintainer burnout is triage, not code. Run a @triage agent on your GitHub webhooks — labels, duplicate detection, missing-repro asks, security flagging — and a @repro agent that attempts minimal reproductions in a scratch checkout. You keep the merge button; Monday brings a stale-sweep digest to your Telegram.

Why this shape

• Triage without takeover. @triage never closes issues; the stale sweep is a digest to you, not an action. First-time reporters get a warmer tone.
• Repro as a service. @repro reports exact steps and environment as comment drafts — it never pushes code or opens PRs.
• Security-aware. Vulnerability-sounding reports are flagged for private handling instead of public replies.
• Scales across repos. One "repo": "*" route covers your whole account; per-repo routes override where a project needs a different agent.

Start

Ready-to-run config: examples/maintainer-fleet/

cp examples/maintainer-fleet/agentx.json agentx.json
# .env: GITHUB_TOKEN (fine-grained, issues:write), GITHUB_WEBHOOK_SECRET, TG_BOT_TOKEN
mkdir -p workspaces/{triage,repro}
agentx daemon start

Point your repos' webhooks (issues, issue_comment, pull_request) at the daemon — receiver + signing setup in journey ch. 5.

Grow

• Heavy repro jobs on a beefier meshed machine → mesh federation
• Triage knowledge compounds (flaky tests, common misconfigs) → shared wiki