`api_integrity_check`

Complete typed causal order existence plus bounded per-cell race, stale-read, and declared Yichus/Spec checks. Inspect the returned claims: top-level ok and WebMCP isError describe tool execution, not whether properties hold. Default schema=legacy-2.2 retains verdicts.invariants; opt in to schema=3.0 for spec-verdict claims with typed subject, operation scope, completion and trust gaps. A declared model hold does not prove its implementation, and a sampled hold is inconclusive in 3.0. Truncated cells without a counterexample remain inconclusive; causal consistency alone does not certify model quality or database executions. TLC is not run; both schema views carry only a diagnostic sketch. For legacy-2.2, view=compact replaces only the inline sketch. For 3.0, it also deduplicates operations into an operationCatalog and uses numeric references in scopes and causal pairs. Both compact views carry a sketch hash and full-view retrieval request; view and schema are independent selectors.

Kind: IntegrityCheck. Origin: Yichus/Mcp::catalog.

CLI: yichus spec — the same native bounded/structural checker; the CLI requires a declared Property or StepProperty, while the MCP tool can inspect built-in invariants alone.

Parameters

sources (required, array) — JSON array of {fileName, source} Bosatsu files.
module_name (optional, string) — Optional module name for the verdict (default Api).
max_ops (optional, integer) — Optional schedule bound (default 8, hard cap 8); does not limit complete causal graph checking.
view (optional, string) — Optional full (default) or compact. Compact retains the decision evidence and a retrieval reference for the diagnostic sketch. Unknown values are errors.
schema (optional, string) — Optional legacy-2.2 (default) or 3.0. Select the evidence schema independently of view; unknown values are errors.

Read the result according to this tool’s scope: static checks, bounded execution checks, and descriptive diagrams answer different questions. A successful call is not a general approval of the program. The safety and permissions guide compares the checks and provides editable ownership, guard, and role examples.