Skip to content

Capability catalog

The twenty original capabilities retain 1.0.0; the nine email capabilities use 4.0.0. Each manifest includes JSON request and response schemas, effects, requirements, examples, projections, and typed error codes.

Capability Input focus Output Effect
creator.research_synthesis@1.0.0 Publication, research window, evidence queries ResearchSynthesis Read
creator.extract_editorial_signals@1.0.0 Normalized source observations Editorial signals Input-only read
creator.update_story_revisions@1.0.0 Signals and prior immutable revisions Story revisions Input-only read
creator.build_story_dossier@1.0.0 One story revision and prior coverage StoryDossier Input-only read
creator.plan_editorial_agenda@1.0.0 Dossiers, desk, coverage window, slots EditorialAgenda Input-only read
creator.plan_edition@1.0.0 Agenda, publication window, update policy EditionPlan Input-only read
creator.plan_content_portfolio@1.0.0 Syntheses, objectives, history, requirements, planning window ContentPortfolioPlan Input-only read
creator.create_content_brief@1.0.0 Assignment, synthesis, profile, creative direction ContentBrief Input-only read
creator.identify_engagement_opportunities@1.0.0 Conversation, persona, evidence, expiry EngagementOpportunity Input-only read
creator.compose_commentary@1.0.0 Opportunity and conversation CommentaryDraft Input-only read
creator.review_commentary@1.0.0 Draft, context, position and experience refs CommentaryReview Input-only read
creator.plan_newsletter_issue@1.0.0 Edition, dossiers, audience policy refs NewsletterIssuePlan Input-only read
creator.compose_newsletter_issue@1.0.0 Accepted issue plan and dossiers HTML/plain-text draft Input-only read
creator.review_newsletter_issue@1.0.0 Plan and draft NewsletterEditorialReview Input-only read
creator.compose_news_article@1.0.0 Article plan and frozen dossiers NewsArticleDraft Input-only read
creator.review_news_article@1.0.0 Plan, draft and qualified-human review ref EditorialReview Input-only read
creator.prepare_correction@1.0.0 Reviewed article and corrected claims CorrectionNotice Input-only read
creator.validate_delivery@1.0.0 Brief, artifact manifest, profile, synthesis, channel plan DeliveryReviewBundle Input-only read
creator.prepare_distribution@1.0.0 Accepted review and exact channel plan Publication proposals Input-only read
creator.assess_performance@1.0.0 Profile, metric queries, objective Performance assessment Read
creator.plan_email_campaign@4.0.0 Supplied email artifacts EmailCampaignPlan Input-only read
creator.finalize_email_campaign@4.0.0 Exact campaign, messages and sequences EmailCampaignRelease Input-only read
creator.propose_email_operation@4.0.0 Exact effect intent and originating evidence ProposedEmailOperation Input-only read
creator.plan_email_sequence@4.0.0 Supplied email artifacts EmailSequencePlan Input-only read
creator.plan_email_message@4.0.0 Supplied email artifacts EmailMessagePlan Input-only read
creator.compose_email_message@4.0.0 Supplied email artifacts EmailMessageDraft Input-only read
creator.review_email_message@4.0.0 Supplied email artifacts EmailEditorialReview Input-only read
creator.prepare_email_delivery@4.0.0 Supplied email artifacts EmailDeliveryPackage Input-only read
creator.assess_email_program@4.0.0 Supplied email artifacts EmailProgramAssessment Input-only read

Use zeo-creator capabilities --json for canonical manifests or zeo-creator capabilities --projection openai for function-tool projections.

Zeocore 0.9 requires every manifest to carry at least one effect kind. Pure transformations therefore conservatively declare read; the email family also carries the pure tag and has no required acquisition services or network access. This compatibility boundary can be removed if Zeocore gains an explicit pure effect.

Workflow composition, scheduling, retries, approval waits, and effect authorization are runtime concerns; the package deliberately exposes no monolithic workflow.

The query-based research and performance v1 capabilities are retained for legacy compatibility. Canonical email assessment uses already-collected observations in assess_email_program; never inject a live connector into that capability.