A.3 Auditable Transformation
A.3.1 The Capability
Make consequential transformations reconstructable and detect silent loss. Record who changed what, then verify that the transformed artifact retained the required semantics.
A.3.2 When This Stack Earns Its Keep
Reach for it where:
- Multiple actors or agents mutate important artifacts, making attribution and transformation history consequential.
- Partial or bypassed instrumentation is plausible, so "we log our writes" is not proof every write was logged.
- Transformation can silently lose product semantics — the output looks fine and is not.
- Incident reconstruction requires both facts at once: who changed what, and whether the result remained faithful.
A.3.3 The Composition
A.3.4 Constituent Moves
| Move | Role |
|---|---|
| MARK | Attach identity and context to the mutation. |
| EMIT | Persist a structured record. |
| COVER | Detect mutations that escaped attribution. |
| READ | Reconstruct the transformation history. |
| FIDELITY | Check that the transformed artifact retained its required semantics. |
A.3.5 Why These Travel Together
Attribution without completeness produces a persuasive but partial history—convincing exactly where it is silent. Completeness without readable provenance proves only that records exist. And provenance alone cannot establish that a transformation preserved the product.
The stack must therefore establish both causal history and product fidelity: How did this artifact reach its current state, and did the transformation preserve its required semantics?
Mechanisms: provenance stamping · attribution coverage · fidelity gate
© James C. Davis, 2026–present