Reason over SourceOS State Integrity failures, conflicts, and repair plans

[mdheller]

Purpose

Holmes should reason over SourceOS State Integrity events and reports to explain causality, recommend safe repair paths, and distinguish policy, schema, conflict, transport, and corruption states.

Canonical spec: SourceOS-Linux/sourceos-spec/docs/architecture/sourceos-state-integrity-layer.md
Program tracker: SourceOS-Linux/sourceos-spec#83

Reasoning targets

• object conflict causality
• schema drift and compatibility failure
• profile boundary violation
• policy-denied replication or write
• degraded derived index
• safe repair vs human-review-required repair
• agent transaction lineage
• memory provenance and invalidation chains

Requirements

• Consume structured events and repair reports rather than raw logs.
• Explain root cause with actor, device, profile, workspace, object, schema, and policy context.
• Produce recommended next actions with confidence and risk classification.
• Do not recommend destructive operations without explicit scoped review.

Acceptance criteria

• Holmes defines a State Integrity reasoning input schema.
• Holmes can generate a causal explanation from conflict, policy denial, and repair event examples.
• Recommendations distinguish safe automatic repair, review-required repair, and blocked-by-policy states.