Mendmark 0.x is evolving, but automation still needs predictable contracts.
- Python 3.10 through 3.13 on Linux are supported. CI tests every supported version. macOS and Windows are not yet claimed as supported platforms; their evaluator command parsing must be validated before making that claim.
- Patch releases preserve CLI flags, report fields, JSON protocol fields, and mutation IDs except where a security defect makes that unsafe.
- Minor releases may add optional object fields. Consumers must ignore unknown
fields and select behavior using
schema_version, not the package version. - JSON suite and evaluator protocol
1.0remain the stable flat-trace contract. Native event graphs use2.0; Mendmark chooses the evaluator protocol from the case type and never sends2.0fields to a1.0single-agent audit. - Removing or changing a field, status, mutation ID, or plugin requirement needs a new schema or protocol version and migration notes.
- Built-in mutation IDs remain
<case_id>:<operator_name>:<stable_suffix>. - Historical benchmarks select immutable
agent-eval-v1ormulti-agent-v1mutation profiles. Ordinary audits usecurrent; adding a new operator never rewrites an older golden-set contract. - Harness adapters are dependency-light and use documented public object fields. Hosted CI exercises LangChain Core 1.5.3, CrewAI 1.15.14, and OpenAI Agents SDK 0.19.4 as the 2026-08-10 compatibility snapshot. Newer supported releases are tested through the latest-version harness matrix; Mendmark does not install or pin a harness in customer environments.
- A custom operator name is globally unique within an audit and is part of the customer's accepted baseline contract.
audit,audit-json,equip,prepare,grade, andshowexit with 0 for success, 1 for a failed product gate, and 2 for invalid input or infrastructure failure.
The JSON Schemas under mendmark/schemas are the machine-readable contract.
Python suites and plugins are trusted-code APIs and receive deprecation notes at
least one minor release before removal when security permits.