← Back to journal

The Output Folder Is Part of the Safety Boundary

Why parity reports should not overwrite unrelated work or become part of their own input.

Evidence basisKalu, Korpus, and frozen Ushman source/docs snapshot reviewed September 17, 2026; no new runtime qualification performed
DisclosureEngineering or methodology note based on a September 17, 2026 snapshot. Source-level behavior is not a public availability or universal compatibility claim. Any scenarios are illustrative, not customer outcomes.
Published by Kalu KodeRequest a KodeBack assessment →

Scope: Engineering or methodology note based on a September 17, 2026 snapshot. Source-level behavior is not a public availability or universal compatibility claim. Any scenarios are illustrative, not customer outcomes.

A reporting command can cause damage without changing application code. It can write inside the tree it is measuring, follow an unexpected path alias, or replace a directory it does not actually own.

KodeBack’s parity publication path treats output placement as a bounded operation rather than an incidental filesystem convenience. That is important for a tool whose reports are supposed to clarify evidence, not alter it accidentally.

Keep the report outside the subject

The current command refuses an output directory inside the candidate or capture directory. Otherwise, producing a report could change the next candidate identity or contaminate the sealed input.

In a synthetic workflow, an engineer might naturally choose a convenient reports folder under the candidate. The refusal makes the ownership conflict explicit before files are written. The report needs a home that is separate from the artifact it describes.

Replace only an owned result tree

Publication stages a bounded output tree and uses an owned destination. A previous KodeBack parity tree can be replaced through that path. An unrelated directory at the destination is refused rather than deleted to make room.

This matters during repeated investigation. A recovery tool should not interpret an arbitrary existing folder as disposable just because the operator selected it as output. Ownership must be established, not inferred from convenience.

The receipt describes publication, not application behavior

A successful report publication means the intended report bytes reached the intended owned location. The report’s actual verdict may still describe a measured difference, a blocked comparison, or an unsupported observation.

Likewise, a publication receipt does not turn the report into mutation authority for a later operation. The later owner still needs to verify current evidence and its own preconditions.

This article describes a source-level boundary in the current parity path. It is not a claim of arbitrary filesystem safety or universal host support. The product remains bounded by its declared environment and supported input forms.

For engineers, the lesson is practical: output handling belongs in the review of evidence tools. A report that changes its own subject or erases unrelated work has already weakened the investigation before anyone reads the verdict.