[drk-ram]

Summary

Closes #101

• Walk (list, tuple) in _prohibited_paths, prohibited_paths, and _json_shape_error in src/agentrust_telemetry/validation.py and conformance/runner/validate.py, and reject duplicate JSON object keys via object_pairs_hook when loading JSON files.
• Reject array-valued and non-string carrier headers with PropagationError in packages/typescript/src/propagation.ts, and check !== undefined instead of truthiness in ExtractedContext.eventFields, matching Python ^[xtract_context.
• Validate �uild_digest / �uildDigest and policy �undle_digest (^Glgorithm and �alue) against canonical lowercase ^sha(256:[0-9a-f]{64}|384:[0-9a-f]{96})$ syntax in both src/agentrust_telemetry/trace_adapter.py and packages/typescript/src/trace-finalizer.ts before signing.

Contract impact

• No wire-contract change
• Backward-compatible contract change with fixtures
• Breaking contract change with migration notes

Evidence

• python tools/sync_schemas.py (passed)
• python tools/check_schemas.py (passed: 8 bundled schemas match normative bytes)
• python conformance/runner/validate.py (passed: all 18 conformance and appraisal fixtures pass)
• python -m unittest discover -s tests -v (passed: 165 tests, 163 passed, 2 Ubuntu-only workflow tests skipped on Windows)
• [x]
  pm --prefix packages/typescript run check (passed: 46 TypeScript tests passed, sc build and
  pm pack --dry-run succeeded)
• python -m build (not run; �uild module not installed in local venv)

Privacy and security

• No real prompts, source, credentials, personal data, or production identifiers added
• New fields have capture semantics and privacy classification
• Security-sensitive changes received maintainer review

Limits

Validates digest syntax and JSON/container shapes in Python and TypeScript SDKs; does not alter the JSON Schema wire contract or cryptographic verification semantics in ^Ggentrust-trace.

[github-actions]

🟡 Contributor Check: MEDIUM

Check	Result
Profile	MEDIUM
Credential	LOW
Overall	MEDIUM

Automated check by AgenTrust Contributor Check.

[dinakarjs]

Reviewed exact head 4f42131 (full commit attached). Checked tuple traversal for privacy/JSON-shape checks, duplicate-member schema/fixture parsing, digest validation and TypeScript propagation/header refusal. With hash-locked Python test dependencies, all 165 unittest cases and conformance validation passed. TypeScript build passed and all 46 tests passed using node --import tsx --test test/*.test.ts (the tsx CLI IPC socket is unavailable here). The first Python run inherited a host 1% OTel sampler; clearing OTEL_TRACES_SAMPLER/ARG resolved the three sampling-dependent failures, which also reproduced on main. Hosted CI/CodeQL/fuzz are still action_required, so this is a source/local-test approval and does not claim hosted validation.

[rajnisht7]

the overall changes looks good, a few things needs to be addressed:

• Digest tests - Python suite tests trailing newline (sha256:…\n). TS regex rejects it too, but the new test matrix doesn’t cover it, add the same case for both buildDigest and policy bundle_digest.
• extractContext wording -The changelog/PR says “carrier headers”, but the code only validates the three AgenTrust headers (run-id, workflow-id, agent-id). OTel headers (traceparent etc.) are still passed through, matching Python, kindly narrow the wording to “AgenTrust carrier headers” so the claim matches the implementation.

[drk-ram]

Thanks @rajnisht7! Added trailing-newline (sha256:...\n) regression cases for both buildDigest and policy.bundle_digest in packages/typescript/test/trace-finalizer.test.ts, and narrowed the PR title and description wording to "AgenTrust carrier headers".