Agent Context Evals v0.11.0 - Fresh-Clone Reproducibility

Agent Context Evals v0.11.0

Status: public fresh-clone reproducibility receipt for the v0.10 local
machine-evidence path. No public benchmark claim.

What Shipped

• Fresh-clone reproducibility receipt:
  release/v0.11-fresh-clone-reproducibility/fresh-clone-receipt.json.
• Manual release-operation generator:
  scripts/run_fresh_clone_reproducibility.py.
• CI-safe receipt gate:
  scripts/check_v11_fresh_clone_receipt.py.
• 5-minute local run feedback request:
  https://github.com/ctxgov/agent-context-evals/issues/22.

Reproduce

python3 scripts/check_v11_fresh_clone_receipt.py
python3 -m unittest tests.test_v11_fresh_clone_reproducibility -v

To regenerate the receipt, run the explicit release-operation path. It uses
network access to clone the public repository into a temporary directory:

python3 scripts/run_fresh_clone_reproducibility.py --ref v0.10.0

The checked-in receipt records a fresh clone of the v0.10.0 public local
evidence path, then runs:

• python3 scripts/check_v10_saved_trace_readiness.py
• python3 -m unittest discover -s tests -v

Boundary

• No public benchmark claim.
• No security claim.
• No provider/model call.
• No adoption claim.
• No human reviewer claim.
• No package publication claim.
• No stable protocol claim.
• No hosted runtime or live adapter claim.

This release publishes a reproducibility receipt and a local receipt gate only.
It does not execute provider/model calls, package publication, hosted runtime
changes, target writes, reviewer outreach, adoption measurement, or public
benchmark reporting.