# Substance — `context` step **Output path:** `.forge/project-context.json` **Topic focus:** construct the machine-readable project context from discovered facts. You run *after* the indexes are written. Describe *the project as structured data*, mapping findings onto the schema fields. **Discovery input to read:** the `x-placeholder` annotations in `{bundleRoot}/schemas/project-context.schema.json` (the field guide) plus all KB docs and Phase-1 findings for the values. **Required output:** - A valid `.forge/project-context.json` matching the schema. - Required fields `project.name` and `project.prefix` must be non-empty strings. - Every array field must be an array (never null or a bare string). - Map only observed facts; mark uncertain values conservatively rather than inventing them. **Not applicable:** never skipped — `project-context.json` is mandatory. A missing or empty `project.name` is a hard Phase-2 validation failure: `× Phase 2 validation failed: project.name is missing or empty.`