[drk-ram]

What this changes

Closes #648.

Hardens conformance/src/trace_tests/loader.py and conformance/src/trace_tests/report.py:

• Rejects non-standard JSON constants (NaN, Infinity, -Infinity) via parse_constant=_refuse_constants and non-bytes inputs (not isinstance(raw, bytes)) in loads_strict with LoadError, preserving loads_strict and parse_record's documented LoadError-only boundary while keeping invalid_canonical_non_finite_float.json (1e400) reachable for TR-SIG.
• Rejects non-dict record inputs in record_digest and build, and validates results_by_level (requiring exact int keys in LEVEL_NAMES and dict values) in build.

Validation commands and results:

• PYTHONPATH=src;conformance/src python -m pytest tests/test_conformance_report_parity.py -q -> passed (26 passed)
• PYTHONPATH=src;conformance/src python -m pytest conformance/tests/test_report.py conformance/tests/test_canonicalization_refusals.py conformance/tests/unit/test_loader.py -q -> passed (51 passed)
• python -m ruff check tests/test_conformance_report_parity.py -> passed (All checks passed!)
• python tools/check_dashes.py -> passed (no dashes outside examples/, spec/trace-v0.1.md)
• PYTHONPATH=src python tools/review_bar.py --base origin/main --head HEAD -> passed (14 changed test cases fail on behaviour at origin/main)

Type of change

• Editorial (typo, link fix, clarification: no normative effect)
• Non-breaking spec change (new optional field, new platform profile, informative addition)
• Breaking spec change (requires 14-day comment period and Project Lead sign-off)
• Schema change
• Example addition

Spec section

conformance/src/trace_tests/loader.py and conformance/src/trace_tests/report.py (conformance test suite loader and report builder; no normative change to spec/trace-v0.2.md)

Tests for a fix

• parse_constant=_refuse_constants and isinstance(raw, bytes) in loads_strict: tests/test_conformance_report_parity.py::test_loads_strict_and_parse_record_reject_non_standard_constants_and_non_bytes
• isinstance(record, dict) in record_digest and build: tests/test_conformance_report_parity.py::test_record_digest_and_report_build_reject_non_dict_record
• results_by_level level and dict validation in build: tests/test_conformance_report_parity.py::test_report_build_rejects_invalid_results_by_level

Changelog entry

• Fixed (conformance(loader,report): reject non-standard JSON constants in loads_strict and validate record_digest/build inputs #648). trace_tests.loader.loads_strict now rejects non-standard JSON constants (NaN, Infinity, -Infinity) and non-bytes inputs with LoadError, and trace_tests.report.record_digest and build reject non-dict records and invalid results_by_level mappings with ValueError.

Checklist

• DCO sign-off on all commits (git commit -s)
• Breaking changes marked with <!-- CHANGED: #NNN: description --> in spec text
• Backward compatibility statement included (for breaking changes)

[github-actions]

🟡 Contributor Check: MEDIUM

Check	Result
Profile	MEDIUM
Credential	LOW
Overall	MEDIUM

Automated check by AgenTrust Contributor Check.

[rajnisht7]

Hello @drk-ram, thank you for the continued interest in contributing to agentrust-io, before we go any further, we need to sort out what's already open.

Quick recap of where things stand across agentrust-io: 120 PRs opened in the last 17 days, averaging over 7 a day. 69 are currently open, 46 merged, 5 closed without merging, and almost 98 PRs landed in earlier 7 days.

You also mentioned in your maintainer interest issue #640 that you review all code diffs and test output yourself before submitting, with AI agents working under your supervision. Given what we're actually seeing in the open PRs, that review step doesn't seem to be happening consistently. Some examples:

• Several PRs have gone through 2-3 rounds of review comments and still aren't merge ready
• A few have failing test coverage that's not addressed yet
• Several have merge conflicts that just sitting there unresolved
• Several fail CI because the branch was not rebased or updated against main yet
• Some PR descriptions and comments have broken characters in them, like �zure_vtpm._b64url_uint &, which is the kind of thing a human proofreading their own PR description would normally catch
• Some already have changes requested that are not addressed yet
• And most of the PR batches landed almost at the same time, which isn't really how one person writing, testing, and reviewing each PR individually would show up

If each of these was reviewed before opening, these issues would have been caught before submission, not after. At this volume and with these gaps, it's hard for us to review each, and also whether it's you behind each PR or something more automated opening them faster than they can reasonably be checked.

So kindly avoid opening new PRs until the current backlog is cleared,kindly go back through your own open PRs and actually review them yourself first, fix the ones with pending changes requested, resolve the conflicts, fix the failing tests and CI, check if there are still bugs or issues which still exists in the diff and fix them and get them into merge ready state

And separately, it would help if you could just clarify directly: are you personally testing, and reviewing each of these before they go up, or is there is an automated process submitting PRs on this account with less oversight than described? We'd rather have a straight answer than keep guessing from the pattern.

[drk-ram]

Thank you @rajnisht7 for the direct feedback — point taken, and I am pausing all new PR submissions across agentrust-io until the existing open PRs are reviewed and cleared.

To answer your question directly: I work on Windows 11 and had been batching my local fixes and submitting them together using a PowerShell gh pr create script. In PowerShell double-quoted strings, a backtick before a, b, e, r, or t (such as `azure_vtpm or `request_sha256) is interpreted as an escape sequence (\a BEL / ^G, \b, \e, \r, \t), which caused the broken characters in weight-custody-manifest#224 and integrations#364. Pushing those batches at once also meant that when upstream main moved (e.g., CHANGELOG.md merges or the pip-audit dependency bumps), multiple open branches conflicted or failed CI at the same time.

I have gone back through every open PR across all 8 repositories to resolve all merge conflicts, fix the PR descriptions, and address every pending review comment:

1. agentrust-io/trace-spec:
   • fix(intent_bridge,successor_observation): validate caller options and JWKs at entry and deep-copy observation (#617) #618 (76cacd8): Rebased onto b938790 and updated _clone_observation to clone tuple containers iteratively (APPROVED by @dinakarjs).
   • prototype(azure_execution_binding): bind HCLAkPub exponent e and key size, reject duplicate JSON keys, and harden _secure_boot_state and authenticode_sha256 #620 (5ab7fbc): Rebased onto b938790 and added the 512-byte zero-section PE image test isolating cert_offset < headers_size (23 passed).
   • fix(trust-record-verify): reject non-object inputs in parentRecordHash, verifyDelegationLink, and bundleDigest (#616) #622 (916cec5): Rebased onto b938790, fixed the vector filename (20-malformed-statement-on-another-log.json) and vec.records[0] in trust-record-verify/tests/units.test.mjs (npm --prefix trust-record-verify test: 165 pass, 0 fail), and expanded tests/test_trust_record_verify_citations_and_platform.py.
   • conformance(inclusion): enforce registry-anchor-v1 surrogate/key/cycle rules and validate leaf_index/entry fields (#625) #630 (b9cc533): Rebased onto b938790 and narrowed parse_entry and spec/registry-anchor-v1.md §4 on producer/batch_id to reject blank strings and ASCII control characters while preserving internal spaces.
   • docs/spec: close Go verifier findings on Anchor Format §1 escaping rules, verifier-compatibility README, and Acta crosswalk (#536) #537 (608c2d6): APPROVED by @noah-ing, @rajnisht7, and @lywinged.
2. agentrust-io/ca2a:
   • Formatted #275 (4af1465), #276 (f231b13), #277 (e7008a1), #278 (8eb696d), #287 (e2fd3a4), and #288 (d9e1939) with ruff format src/ tests/ per @carloshvp's review (#270, #289, #290, #291 are already APPROVED).
3. agentrust-io/cmcp:
   • #805 (5d057eb): Added TPMS_ATTEST type == 0x8018 (TPM_ST_ATTEST_QUOTE) check, required ak.get("kty") == "RSA", and added unit tests for 100% patch coverage.
   • #806 (d897245): Enforced len(report_data_hex) <= 128 in both direct and quote TDX helpers and added 130-char negative regressions.
   • #807 (39fea26): Rebased onto origin/main (1fbc1b2) to resolve the tests/unit/test_policy_bundle.py conflict and added strict UTF-8 / duplicate-key / non-finite validation to _approval_schema().
   • #808 (4ab488a), #816 (f06f3ff), #817 (51c0d56), #818 (417e073): Rebased onto origin/main (1fbc1b2) so they pick up the pip-audit fix.
   • #819 (e4cc625): Added finite parse_float validation on EAR payloads and top-level/nested overflow tests (#800 and #815 are already APPROVED).
4. agentrust-io/agent-manifest:
   • refactor(adapters): share common session validation helpers #552 (4e485ce): Rebased onto a2f50b2 and moved _expected_audit_path_length and range checks ahead of element iteration in verify_inclusion.
   • conformance: expected policy hash as verifier context for TR-POL-003 (#280) #553 (f42bce7): Rebased onto a2f50b2 and made _reject_cbor_sentinels cycle-safe for CBOR value-sharing tags 28/29 (#538 is APPROVED).
5. agentrust-io/weight-custody-manifest, trace-registry, agentrust-telemetry, integrations:
   • Rebased all 9 open weight-custody-manifest PRs onto 8ae4da6 and fixed the escaped characters in the PR descriptions of weight-custody-manifest#224 and integrations#364 (all 5 open integrations PRs are now APPROVED).

I will hold off on opening any new PRs until the current queue is worked through.