Skip to content

Latest commit

 

History

History
21 lines (16 loc) · 1.29 KB

File metadata and controls

21 lines (16 loc) · 1.29 KB

Validation plan — v0.1.0

A release is not considered validated until all gates pass.

  1. Unit tests — hashing/privacy, per-run overrides, policy semantics, review helper, failed tools, fail-open behavior.
  2. Native Strands runtime test — real Agent, native Plugin, native hooks, native decorated tool, deterministic custom Model; no network/model provider.
  3. Python compile check.
  4. Local LoopGrid Core 0.8.1-design-partner E2E with the canonical sandbox demo.
  5. Expected E2E result: lifecycle evidence_complete, evidence coverage 100, verification valid=true, failures [].
  6. GitHub Actions green before tag.
  7. PyPI publication from the tagged source.
  8. Strands integrations catalog PR only after the package/repo are public.

Do not claim Strands/AWS verification, endorsement, partner status, Featured status, or badge unless the Strands maintainers explicitly grant it.

Strands Python invocation-state note

invocation_state is the request-scoped channel consumed by native Strands hooks and tools. Do not read the LoopGrid run_id back from AgentResult.state; current Python Strands uses that field for its event-loop/request-state slot. Keep the caller-owned run_id explicitly and use it with LoopGridPlugin.observe_outcome(...). The runtime test covers this behavior.