Sync Boatstack from Intelligence Flow Labs @ cb079d704e3b

[operator-stack-publisher]

What this sync releases

Boundary Failure Mode Analysis in Auto-Plan

Introduced an opt-in boundary_analysis configuration to .boatstack-project.json. When enabled, the auto-plan operation actively analyzes requested changes to determine if they are symptoms of missing systemic boundaries (e.g., deficient data normalization or leaky validation). Instead of silently expanding scope, Boatstack presents this as a material product decision with concrete codebase evidence and offers tiered implementation paths: a local symptom patch or a programmatic enforcement lock. If programmatic enforcement is selected, the plan is explicitly split into two delivery slices, ensuring the boundary (hook, trigger, or strict test) is established before the feature is implemented.

Deterministic enforcement of Boundary-Oracle Loop

Transformed the Boundary Failure Mode Analysis from a prompt suggestion into a strict, programmatic intelligence harness:

1. Schema Contract: Added systemic_boundaries array to the plan.md schema, requiring each boundary to define a verification_oracle.
2. Structural Lock: Updated ValidatePlan to physically reject plans that define a systemic boundary but fail to provide an oracle or isolate the boundary into its own delivery slice.
3. Execution Lock: Updated /test-gate instructions to require test evidence proving that the verification oracle (negative test) successfully blocks or normalizes a violation attempt.
4. Compounding Memory: Upon PR publication (publish-pr), Boatstack now extracts verified systemic_boundaries from the feature lock and appends them to .product-loop/verified-boundaries.md. Agent system prompts have been updated to read this file and natively respect established boundaries in future runs.

Projection provenance

Generated from operatorstack/intelligence-flow@cb079d704e3b0d82fa88b4318d3fb89fbd96723e.
Review provenance, tests, examples, and context-cost changes before merging.