fix(agt-adapter): appraisal.status is the verifier's, so default it to none

[imran-siddique]

Closes #331.

build_trust_record() set appraisal.status to "affirming" on every record it produced, and __init__ had no parameter to change it. appraisal.status is verifier-owned: section 3.3.1 says a verifier MUST record the depth it actually checked and MUST set the status to contraindicated when evidence fails, and models.Appraisal carries the same point in a comment, "What this verifier ran, not what the issuer claimed."

The record signed and verified, so a consumer reading that field to find out whether anybody checked was told yes by a record nobody had appraised.

Change

appraisal_status becomes a constructor parameter defaulting to "none", mirroring TraceSandboxAdapter, which already had this right with tests pinning both the default and the override. agt.py:163 was the only remaining hardcoded status in the package.

This changes the content of records this adapter emits. An unappraised record now says none where it used to say affirming. That is the correct direction, and a caller who really did appraise declares it with appraisal_status="affirming".

Docs that described the old behaviour

• docs/integration/agt.md said the adapter "populates an affirming appraisal without independently evaluating the session".
• docs/tutorials/agt-adapter.md carried a post-hoc record["appraisal"]["status"] = "none" line, so the gap was known well enough to be worked around in a tutorial rather than fixed in the adapter. That line is gone because the default now does it, and tests/test_docs_quickstart.py executes that tutorial, so the edit is checked rather than asserted.

Verified in both directions

• With the fix: tests/test_agt_adapter.py 25 passed, docs quickstart 2 passed, repository 1444 passed.
• Against origin/main's agt.py: 7 of the new cases fail.
• Coverage is the default, the override, every status the model allows reaching the record and validating as a TrustRecord, and a control that an unappraised record still signs and verifies so the new default costs a caller nothing.
• The four repository-wide failures in test_generators_reproduce_fixtures.py and test_safe_integer_range.py are pre-existing Windows path-separator artifacts.

Reported by @Yatsuiii, who found it by comparing the two adapters rather than by reading either one alone.

🤖 Generated with Claude Code

https://claude.ai/code/session_01X27MBo6tGVw1UmmchuGFnW