feat(experiments): Claim 5 temporal adjacency + Claim 6 cross-org attestation#352
Merged
Merged
Conversation
…rg attestation Claim 5 (temporal-adjacency): 6 properties proven -- monotonic sequence numbers, cross-boundary event detection from phi/pii/pci/restricted domains, provenance disclaimer in every call graph summary, zero false negatives by construction, concurrent call ordering, denied calls recorded. 9 pytest tests. Closes #350. Claim 6 (cross-org-attestation): software simulation of dual-TEE B2B protocol. 7 properties proven -- independent keypairs, session_id linkage, Phase 1 and Phase 2 nonce binding (SHA-256(key||session_id)), independent verification, cross-claim tamper independence, server binary swap detection. 9 pytest tests. Closes #351. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- PLC2701: add noqa for private _HIGH_SENSITIVITY_DOMAINS import - C416: replace unnecessary list comprehension with list() - F841: remove unused variables sv_key, session_id, key Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
… job agent-governance-toolkit-core is already installed as a core dependency via pip install -e ".[dev]", so agt is available without the extra arg. agent-compliance does not exist on PyPI and has been failing CI since #346. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
The agt CLI is provided by agent-governance-toolkit-compliance, not agent-compliance. The latter name does not exist on PyPI; this was the root cause of the governance job failure since #346. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
…st to PyPI - governance job: agent-governance-toolkit>=4.1 is the published meta-package that includes the agt CLI (was using wrong sub-package name) - pyproject.toml: drop git source pin for agent-manifest, use PyPI >=0.1.1 - remove allow-direct-references hatch flag (no more git deps) Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Adds experiments and CI tests for Claims 5 and 6.
Closes #350, #351
Claim 5 — Temporal Adjacency (
experiments/claim5-temporal-adjacency/)The
SessionCallLogtracks calls in monotonically increasing sequence order. Any call B after a sensitive call A has seq(B) > seq(A) — an implicit edge exists. This guarantees zero false negatives: if the agent used A's data in B, the model has a record of the relationship.edges_representfield)9 pytest tests in
tests/unit/test_claim5_temporal_adjacency.py.Claim 6 — Cross-Org Attestation (
experiments/claim6-cross-org-attestation/)Software simulation of the dual-TEE B2B protocol: Phase 1 (enterprise gateway) and Phase 2 (vendor server stub) each generate independent signed TRACE Claims linked by shared
session_id. A third-party verifier confirms both independently.9 pytest tests in
tests/unit/test_claim6_cross_org_attestation.py.Note: P4 (nonce check) is mathematical in software mode. In hardware TEE mode, the nonce is hardware-signed by the TEE provider and verified against the endorsement chain.
Experiment index
Updated
experiments/README.mdto list all 7 experiments (Claims 1–6, Claim 2 has two experiments) with their key results.Test plan
python -m pytest tests/unit/test_claim5_temporal_adjacency.py tests/unit/test_claim6_cross_org_attestation.py -vpython experiments/claim5-temporal-adjacency/run.py && python experiments/claim6-cross-org-attestation/run.py