oxygen-agent.com

oxygen-sequencer

Operate Sequences for net-new LinkedIn and multistep outreach across email, LinkedIn, WhatsApp, call, and CRM tasks. Covers readiness, cadence, enrollment, suppression, reply-stop, preview, dispatch, and recovery. Replies/direct email use Messages.

First seen Jul 3, 2026

Installation

$ npx skills add https://oxygen-agent.com

Similar popular skills

Related neighbors and high-traction skills in the same topics — useful to compare before installing.

Also in this package

Other skills from oxygen-agent.com · top by installs.

npx skills add https://oxygen-agent.com

Browse all from oxygen-agent.com

More details

Agent compatibility

Declared targets from SKILL.md / docs. Unmarked agents are not listed — the skill may still install via the CLI.

Claude Code Not declared
Cursor Not declared
Codex Not declared
GitHub Copilot Not declared
Windsurf Not declared
Gemini CLI Not declared
Cline Not declared
OpenCode Not declared

Skill metadata

Parsed from SKILL.md frontmatter.

Allowed toolsBash(oxygen *), Bash(oxygen-dev *), Bash(jq *)

Package contents

Files included with this skill beyond the listing page.

  • skill md SKILL.md 5,159 B

History

  1. First seen on skills.sh
  2. First recorded snapshot · 73 installs

SKILL.md

OXYGEN Sequencer

Sequences own outreach initiation and cadence, including sender binding, pacing, suppression, reply-stop, retries, provider IDs, and provenance.

Boundary

  • New LinkedIn conversation, even one person/one message: Sequence. The one-step gateway is sequences send / oxygensequencessend.
  • Multi-recipient or multistep email, LinkedIn, WhatsApp, call, or CRM-task outreach: Sequence.
  • Reply in an existing email/LinkedIn/WhatsApp thread: Messages/Unibox.
  • New direct email with no cadence: Messages via email send.
  • Deterministic automation around a campaign: Workflow outside, Sequence inside. Workflows may enroll any Sequence status; draft/paused recipients stay pending. They never send raw provider messages.

Confirm this route with oxygen capabilities search "<outreach goal>" --json, then hydrate the selected command with oxygen commands get or MCP schema discovery.

Objects to inspect

  • Sender pool: accounts/mailboxes, health, warmup, channel limits, and stable person-to-sender binding. Prefer sender profiles (one person's accounts/inboxes, name, and photo), so each lead hears from the same human across channels. Profiles also supply the durable identity for managed-inboxes --sender; never guess names or use a temporary photo (oxygen-email-infra).
  • Definition/version: ordered steps, delays, variants, conditions, tracking, send windows, and channel fallbacks.
  • Enrollment: recipient identity, mapped variables, exclusions, suppressions, prior-contact rules, and tags.
  • Dispatch policy: caps, ramp, rotation/failover, ESP matching, bounce/reply-stop, and the visible footer. includeunsubscribelink is opt-in; absent means off.
  • Run state: previews, sends, skips, failures, replies, costs, and reconciliation evidence.

pause/resume is reversible. Permanently retire only a paused Sequence: delete previews; repeat with --approved. CLI creates no archive; legacy archived rows stay readable/taggable.

HubSpot saved-list enrollment

For recurring HubSpot segment enrollment use sequences hubspot-list-import lists|configure (MCP oxygensequenceshubspotlists / ...hubspotlistimport). Preview first; live needs approval plus its fingerprint. The Workflow applies complete DNC before enrollment and keeps draft/paused recipients pending. Never recreate paging, storage, DNC, or enrollment locally — use the HubSpot runbook.

Performance attribution

Native-email campaign facts:

  • sequences analytics --range 30d --json: all campaigns plus analytics.emailAttribution.byMailbox, .byDomain, and .unattributed.
  • sequences analytics --sequence <sequence> --json: sender/copy performance; unattributed replies stay in campaign totals. sequences contacts <sequence> --contact-state <stage> --json: variables, sender, status, and highest engagement stage.
  • sequences variants <sequence> --json: generic action/base rollup, not copy-only performance. sequences events <sequence> --json: per-lead ledger with mailboxEmail/sendingDomain when preserved.

Mailbox/domain rows report sent, replied, positive, bounced, failed, rates, reply types, and campaign counts. Facts without preserved sender identity stay unattributed; never infer a mailbox from the configured pool.

This is narrower than Messages: messages stats covers Unibox conversations; mailboxes get/domains get|list cover infrastructure. Direct or provider-owned sends gain campaign/mailbox attribution only after telemetry reaches Oxygen's native action/message ledgers, not merely when an integration is connected.

Safe run loop

  1. Inspect channel and sender readiness; do not treat “connected” as “safe to send.”
  2. Create or revise the hosted Sequence definition.
  3. Preview enrollment and dispatch with the exact recipient scope, sender pool, exclusions, credit estimate, and includeunsubscribelink posture (absent = off).
  4. Show both ceilings — the Sequence's lifetime budget and the current live-run cap; neither substitutes for the other.
  5. Obtain explicit approval for the external write and bounded spend.
  6. Enroll, then start. Keep the returned run and Sequence deep-links.
  7. Monitor failures, stops, and sender health. Reconcile ambiguous effects; use sequences recover-capacity only for its previewed, definitive no-effect scope.

Stale scheduling deferrals: pause, preview sequences retry-deferred, apply with --approved, then resume. Email actions refresh their saved window; recovery sends nothing or bypasses window, spacing, quota, suppression, or cap checks.

Never bypass Sequences with raw LinkedIn chat/provider actions, local loops, or a generic integration action. Never resend an ambiguous LinkedIn action until reconciliation proves the provider did not accept it.