Compatibility & protocols · Verify

Compatibility Golden Master Check

Verify that the candidate preserves the selected reference behavior across the supported input and version corpus.

MCP library 1.6.1 · Instruction library 1.6.1 · English technical instructions

← Back to search

Use this workflow

Connect Vibe Coding Library to a coding agent that supports MCP over Streamable HTTP. Choose Other / MCP agent for a neutral prompt, or a ready-made Codex or Claude Code invocation. Add your task details below.

Goal

Verify that the candidate preserves the selected reference behavior across the supported input and version corpus.

Context

Identify authorized reference and candidate versions/configurations, accepted input corpus, expected semantic behavior and fields whose nondeterminism is proven.

Scenarios

Select the cases that test the requested guarantee. Define expected state/effects before execution; use the authorized environment and a valid control.

  • Replay a versioned corpus against reference and candidate; compare semantic output, state, errors, ordering, and side effects.
  • Mutate boundaries and malformed/unknown fields safely; verify parser limits and reference-compatible rejection/recovery.
  • Run duplicate, concurrent, timeout, and restart cases to expose hidden identity, retry, persistence, and ordering contracts.
  • Test multiple reference versions/configurations and classify stable contract versus version-specific behavior.
  • Review every normalized golden field and prove why it is nondeterministic rather than masking a regression.

Verification

Record each reference/candidate mismatch with input identity, output/state/error difference and allowed normalization; do not count a masked semantic difference as compatible.

Execute and decide

Freeze the corpus, reference/candidate versions and normalization rules before comparison. Establish a valid control against both sides, then replay the selected normal, edge and failure cases while observing outputs and state/effects independently. Where order or timing is part of the contract, preserve that observation instead of sorting or dropping it.

Classify each difference as a supported mismatch, proven nondeterminism, environment variance or unobserved behavior. Return passed, failed or blocked for the stated corpus/window. Feed a failing case and its raw/normalized difference back to the authorized repair, repeat the comparison after correction and require Patch Checking for changed code. Regenerating goldens requires reference evidence or an explicit approved contract change.

Workflow ID: compatibility-golden-master-check · View the versioned source · Shared workflow and authority rules