oxygen-gtm
Route an OXYGEN GTM outcome to its native hosted primitive: Knowledge and Recipes; Records, Tables, Messages, Posts, Signals, Dashboards, Ads, and Tags; Sequences, Publishing, Workflows, Agents, and Observability. Use as the fallback router when no narrower OXYGEN skill matches, especially for sourcing, AI/tool/waterfall columns, enrichment, or cross-primitive work.
How do I install this agent skill?
npx skills add https://oxygen-agent.com --skill oxygen-gtmIs this agent skill safe to install?
No partner audit is available yet. Read the source before installing.
What does this agent skill do?
OXYGEN GTM
This is the fallback router, not a command encyclopedia. Prefer a narrower installed skill when the capability card names one.
Discover, then hydrate
Start from the outcome, not a product noun:
oxygen capabilities search "<goal>" --json
oxygen commands get "<recommended command>" --json
With MCP: oxygen_capabilities_search, then oxygen_capabilities_schema for the selected names. Provider operations are a separate catalog: bounded oxygen tools search "<intent>" --json, then oxygen tools get <tool_id> --json. Search selects; get/schema explains exact inputs, current cost, availability, and gates.
Ownership map
| Intent | Native owner | Important negative boundary |
|---|---|---|
| company context, ICP, positioning, rubrics, durable findings | Knowledge Graph | not local chat memory |
| proven business-case playbook | Recipes | advisory; not an execution runtime |
| canonical companies, people, deals, activities | Records | working lists stay in Tables until promoted |
| row work, formulas, AI columns, tool columns, waterfalls | Tables | never rebuild execution in a local spreadsheet/script |
| direct email or reply in an existing thread | Messages | net-new LinkedIn initiation belongs to Sequences |
| post artifacts, comments, reactions, performance | Posts | Publishing owns scheduled delivery |
| durable pulse, engagement, and product events | Signals | not private webhook state |
| curated funnel, spend, and needs-action reporting | Dashboards | not a second analytics store |
| competitor ad intelligence | Ads | owned-audience activation is not implied |
| cross-primitive labels and campaign handles | Tags | not a separate campaign entity |
| cadence, enrollment, suppression, reply-stop, net-new LinkedIn | Sequences | not generic automation |
| scheduled, approval-gated broadcast delivery | Publishing | not one-to-one outreach |
| deterministic triggers, branches, retries, approvals | Workflows | hosted runtime; local files only author definitions |
| adaptive governed goals, threads, checkpoints | Agents | deterministic graphs belong to Workflows |
| runs, costs, failures, approvals, provenance | Observability | read/control lens, not another executor |
No motion running yet? Name the first play, not a primitive: oxygen recipes show linkedin-network-first-motion --json.
Tables, AI columns, and waterfalls
Use Tables for composable row work: describe the table first, then add the column type that matches the job.
- Deterministic transformation: formula column.
- Reasoning, classification, extraction, or drafting: AI column, grounded by workspace Knowledge by default.
- One provider operation: tool/enrichment column.
- Ordered provider fallback for one capability: native waterfall/enrich column.
- Cross-table truth: a Tables-owned relation/link column; linking plain tables never registers CRM objects.
- Multi-step deterministic orchestration: hosted Workflow, not a local loop.
Read guides/cost-precheck.md before a paid pilot. Exact provider details live under provider-runbooks/; load only the selected provider's file.
Native LinkedIn routing
Public or third-party LinkedIn research uses the native cookieless scraper.* catalog by default and consumes no connected-account quota; connected linkedin.* operations are for the workspace account's own private/network data, inbox, and writes. Never silently fall back from the native scraper to a more expensive provider. Read provider-runbooks/scraper.md for the ownership rule and the durable engager-harvest gateway.
Operating loop
- Confirm the active workspace and resolve relevant Knowledge context.
- Search the capability catalog and accept the native owner unless the user chooses otherwise.
- Hydrate one command/tool schema and inspect readiness.
- Preview the smallest representative scope. For Tables, start with a few rows.
- Before credits or external writes, show scope, estimated credits, and side effects; require explicit approval and a hard ceiling. Stop with
AWAIT_APPROVALuntil that approval arrives. - Execute on OXYGEN's hosted primitive. Long work becomes a durable run.
- Return the
web_url, cost/provenance, and any retryable failures.
Focused guides:
- context-bootstrap.md — workspace context.
- finding-companies-and-contacts.md — sourcing into Tables.
- enriching-and-researching.md — AI/tool/waterfall columns.
- writing-outreach.md — grounded copy and qualification.
- workflow-builder.md — deterministic orchestration.
- operating-runs-and-provenance.md — runs, failures, costs, retries.
The live search/get and schema surfaces are the source of exact grammar.
How can the creator link this skill?
Add the canonical catalog link to the repository README so users can inspect current installs and available audits. The publishing guide covers the complete discovery path.
<a href="https://skillzs.dev/skills/oxygen-agent.com/oxygen-gtm">View oxygen-gtm on skillZs</a>