Agent skills for interacting with the Lytics Customer Data Platform. Compatible with skills.sh.
Install every skill for one agent, non-interactively:
npx skills add lytics/agent-skills --skill '*' --agent claude-code -y- Repeat
--agentfor more than one agent (--agent claude-code --agent cursor); a comma-separated list is rejected. Without--agent, the CLI asks which agents to install into, and the "Universal (.agents/skills) -- always included" group it shows is a list of agents, not extra skills. --skilltakes skill names (--skill lytics-audiences --skill lytics-schema) or'*'.- Plain
npx skills add lytics/agent-skillswalks through both choices interactively.
Each skill is self-contained: the shared files it relies on (auth, confirmation gate, FilterQL grammar, ...) ship inside its own references/ folder.
0.4.0 merged 24 skills into 8 and renamed them all with a lytics- prefix. Installing the new ones does not remove the old ones, and two generations side by side will compete for the same requests, so remove the old names first:
npx skills remove audience-builder audience-advisor audience-snapshot segment-manager filterql-builder \
entity-lookup profile-explorer profile-investigator \
integration-advisor integration-setup connection-manager job-manager \
schema-discovery schema-manager schema-optimizer data-health-monitor stream-inspector \
flow-manager campaign-flow-builder export-debugger webhook-template-builder \
account-sync content-affinity-advisor lytics-agent -y
npx skills add lytics/agent-skills --skill '*' --agent claude-code -yAdd -g to remove if you installed globally.
The top-level references/ folder is the only copy to edit. Skills link these files as references/<file>.md. After changing one, or linking a new one from a skill, run scripts/sync-references.sh to refresh each skill's copy. CI fails a PR whose copies are out of date.
| Variable | Required | Description |
|---|---|---|
LYTICS_API_TOKEN |
Yes | Lytics API token for authentication (create one) |
LYTICS_API_URL |
No | Custom API base URL (defaults to Lytics production API) |
See references/auth.md for the full authentication contract, including multi-account, SaaS, and multi-agent deployment contexts.
Each skill covers one area and has modes; its SKILL.md says which file to read for each.
| Skill | Covers | Modes | Replaces (before 0.4.0) |
|---|---|---|---|
lytics-audiences |
Audience segments and FilterQL | build, advise, snapshot, manage | audience-builder, audience-advisor, audience-snapshot, segment-manager, filterql-builder |
lytics-profiles |
Individual profiles | lookup, explore, investigate | entity-lookup, profile-explorer, profile-investigator |
lytics-integrations |
Providers, auth, connections, jobs, webhook templates, export debugging | advise, setup, connections, jobs, webhook-templates, export-debug | integration-advisor, integration-setup, connection-manager, job-manager, webhook-template-builder, export-debugger |
lytics-schema |
Fields, mappings, identity config, patches | discover, manage, optimize | schema-discovery, schema-manager, schema-optimizer |
lytics-data-health |
Is data flowing; streams | health-check, streams | data-health-monitor, stream-inspector |
lytics-flows |
Flows / journeys | build, manage | campaign-flow-builder, flow-manager |
lytics-content |
Content work: topics, Supertopics, content affinity | assess, advise, execute | content-affinity-advisor |
lytics-account |
Account management: users and roles, API tokens, account settings, and copying metadata/settings between accounts (sandbox -> prod). Writes and cross-account sync run only when explicitly asked | users, tokens, settings, sync (compare, resume) | account-sync |
There is no router skill any more (lytics-agent was removed): agents pick a skill from its description, so each description says when it applies.