Skip to content

ARTIFACT_CONTRACTS.md

Latest commit

 

History

History
32 lines (26 loc) · 1.78 KB

ARTIFACT_CONTRACTS.md

File metadata and controls

32 lines (26 loc) · 1.78 KB
Raw

Artifact Contracts

The Skill-RUP framework establishes strict, deterministic contracts for every artifact generated or consumed during the AI lifecycle.

Machine-Readable States

All deterministic state is preserved in JSON format conforming to strict JSON Schemas (Draft 2020-12):

  • RUP_DISCOVERY.json: Output of discovery phase. Governed by schemas/discovery.schema.json.
  • RUP_PLAN.json: Output of the planning phase. Governed by schemas/plan.schema.json.
  • RUP_EXECUTION.json: Output of the execution phase. Governed by schemas/execution.schema.json.
  • RUP_VERIFICATION.json: Verification and testing outcomes. Governed by schemas/verification.schema.json.

Skill-only sidecars (not part of the canonical contract) live alongside the canonical artifacts and are validated against schemas/rup-schema-derived.schema.json:

  • plan-state.json: planning constraints, escalations, dependency-closure admissions, and the per-workstream checkpoint graph.
  • execution-state.json: per-item dispositions and checkpoints, content baseline (HEAD + per-path hashes), content-addressed backups, package change grouping, and the single platform-neutral rollback operation list consumed by both reporting and the rollback CLI phase.

Human-Readable Reports

  • Markdown equivalent reports (RUP_DISCOVERY.md, RUP_PLAN.md, etc.) are synthesized from JSON state for human review.
  • These reports provide diffs, rationale, and execution context.

Schema Validation

At any point, the integrity of an artifact can be validated via validate_rup.py:

python scripts/validate_rup.py --schema protocol/rup-schema.json output RUP_DISCOVERY.json discovery

Failure to conform to the schema constitutes an automatic pipeline failure in forward_test.py and the validate-skill GitHub workflow.