AI Research Preflight v0.1.0

First release of AI Research Preflight.

Includes:

• plan-research-compute: a provider-neutral, no-launch preflight for sanitized research-compute manifests, governed data metadata, abstract DAGs, pinned tools, and bounded resources.
• check-formulation-readiness: a local, redacted checker for user-supplied formula, policy, evidence, arithmetic, units, declared bounds, process order, pair coverage, and experimental gates.

Verified before release:

• 112 local unit and adversarial tests on Python 3.11 and 3.14.
• GitHub Actions passed on Python 3.10, 3.12, and 3.13.
• Current-tree and full-history public-scope checks passed.
• The exact 34-file source archive passed a no-extraction audit.
• Both Skills were previewed and clean-installed from the v0.1.0 tag; installed engine hashes matched the audited source.

Boundaries:

• No raw-data access, cloud submission, workflow execution, network client, subprocess, telemetry, or hidden upload.
• No formula generation or optimization, undeclared compatibility inference, wet-lab instructions, or scientific conclusions.
• A ready result means only that supplied metadata passes supplied rules; it does not establish feasibility, safety, stability, efficacy, compliance, cost, or success.