v2.2.0 Audit Closure and Stage Handoff Edition

CHARMM-GUI System Builder v2.2.0

v2.2.0 adds a formal, fail-closed closure between a completed CHARMM-GUI build and any later MD workflow. It preserves the v2.1 guided decision and build-contract workflow while adding immutable Stage 2 evidence and handoff controls.

Major upgrades

• Vendored simulation-stage-contracts v1.0.0 with deterministic file and tree-hash verification.
• Build contract schema 2.2 with non-destructive v2.1 compatibility.
• Evidence dependency and freshness checks that invalidate stale downstream reports.
• Semantic topology and segment reconciliation instead of literal file equality.
• Capability-aware strict grompp plus TPR readback; no -maxwarn bypass.
• Structural-environment parity ledger for resolved lipids, ions, cofactors, and pocket components.
• Pose-preservation evidence with explicit protein/pocket alignment scope.
• Restraint file, macro, equilibration schedule, and release-handoff validation.
• Receipt 2.0 read-only audit, separately authorized atomic lock, independent verification, and hash-chained supersession.
• Public synthetic fixtures replace the historical target-specific case summary.
• Explicit PyYAML and jsonschema runtime dependencies, tested on Python 3.11, 3.12, and 3.13.

Validation

• 130 unit tests passed on Python 3.11, 3.12, and 3.13.
• Agent Skills reference validation passed.
• Package validator and vendored shared-core integrity checks passed.
• README and installation documentation updated for v2.2.0.

Scientific and execution boundary

Every Stage 2 receipt remains production_ready=false, md_execution_allowed=false, and no_mdrun=true. This release does not run MD, approve production, prove a ligand pose, or establish a biological mechanism.

See the README and CHANGELOG for the full migration and evidence model.

v2.1.0 CHARMM-GUI System Builder

CHARMM-GUI System Builder v2.1.0

This release upgrades the project into an auditable, contract-driven system-building workflow.

Highlights

• Guided Decision Protocol with a complete parameter inventory before builder actions.
• Routine, Contextual, and Critical risk handling with explicit stop conditions.
• Reusable target profiles, SHA-256 locked build contracts, approval/evidence ledgers, and scoped execution authorization.
• Machine-readable official API capability registry and gated Quick Bilayer/status/download client.
• Audited browser-assisted workflow for interactive builders not covered by documented APIs.
• Optional OS keychain/keyring credential broker with memory-only API tokens.
• Versioned rule packs for PDB Reader, Ligand Reader, Membrane Builder, Solution Builder, Quick Bilayer, and GROMACS.
• Stronger archive, package, custom-ligand, and strict GROMACS preprocessing gates.
• Cross-agent support for Codex, Claude Code, OpenClaw, Hermes, and generic Agent Skills clients.

Validation

• 109/109 repository tests passed.
• Python 3.11, 3.12, and 3.13 GitHub Actions checks passed.
• Agent Skills specification check passed.
• Package validator passed.

Safety and scientific boundaries

• production_ready=false
• no_mdrun=true
• No live CHARMM-GUI submission is implied by this release.
• Production use still requires input validation, expert approval, membrane-orientation review, ligand parameter review where applicable, and strict GROMACS preprocessing.

See CHANGELOG.md and the v2.1.0 design documentation for full details.

v1.1.1 Maintenance Release

Maintenance update

• Simplifies duplicate optional JSON metadata loading without changing validator behavior.
• Adds regression coverage for malformed compatibility and provenance metadata.
• Keeps the Cross-Agent workflow and all scientific safety gates unchanged.

Validation: 43 unit tests passed locally; GitHub Actions passed Agent Skills validation and Python 3.11, 3.12, and 3.13 test jobs.

Scientific status remains non-production by default: production_ready=false and no_mdrun=true.

v1.1.0 Cross-Agent Edition

Release sequence

This is the Cross-Agent Edition following the initial v1.0.0 release. The V6 name remains only as the internal workflow/schema identifier.

Included

• one Agent Skills-compatible core
• Codex, Claude Code, OpenClaw, Hermes, and generic runtime adapters
• platform installation guides and capability matrix
• 42 regression tests and official skills-ref validation

Boundary

No scientific behavior changed. production_ready=false, no_mdrun=true, human authentication, and expert approval gates remain enforced.

CHARMM-GUI System Builder v1.0.0

Initial public release of the unofficial CHARMM-GUI System Builder.

Highlights:

• auditable PDB Reader, Ligand Reader, Membrane Builder, Solution Builder, and GROMACS workflow guidance
• resilient Chrome/Safari download recovery and content-based tar/tar.gz/HTML/partial classification
• GROMACS package validation and custom ligand parameter injection checks
• AGPLv3 licensing, transparent Liao Chenyan authorship, canonical provenance, governance, and CODEOWNERS
• 35 standard-library tests across Python 3.11, 3.12, and 3.13

Scientific boundary: a technical validation pass is not production approval or binding-site evidence. No gmx mdrun is performed by this project.