curatelabs/graphforge/graphforge-build-knowledge

Build knowledge with GraphForge

Use this skill when a user wants to synchronize declared definitions, validate an infrastructure target, or move a portable GraphForge project envelope.

Procedure

1. Read .graphforge/graphforge.yaml and validate it with

graphforge config validate.

1. Resolve configuration with graphforge config resolve --json; treat secret

references as references and never print or copy secret values.

1. Run graphforge sync only for definitions and digest-identified external

sources explicitly declared by the configuration. Never scan the repository for data.

1. Before provisioning, run graphforge infra validate --target <name>.

Static validity is not service readiness.

1. Use graphforge export and graphforge import only for versioned portable

whole-project envelopes. Do not treat them as ontology-document export commands. Rust-owned runtime-catalog inspection, deterministic ontology suggestion, non-mutating ontology validation, and explicit YAML/JSON ontology-document export belong to #236; thin Python and Node parity, including durable adopt/clear behavior, belongs to #237.

Safety

Actual graph or source data must remain outside the code repository. Never stage .graphforge/state/, .graphforge/imports/, or .graphforge/exports/. Preserve user-edited or unrelated project skills. Use a checkpoint before destructive lifecycle work, and use graphforge revert when the user requests restoration; revert publishes a new complete generation. Never infer, adopt, clear, or change ontology authority as a side effect of repository sync, project import/export, configuration validation, or IaC.