Repository navigation
Conversation
… gates The 14 PRs merged on 2026-10-16 (#5-#18) landed while GitHub Actions were blocked by billing, so none of them ever ran Lint & Format. ruff check (45 findings) and ruff format (26 files) now pass. Five sources whose exact Git blob IDs are pinned in docs/research/* and evidence/dopa_signal_instrument/* are added to extend-exclude, following the existing dopa_closure_v2.py precedent, instead of being rewritten. Changed files: AST identical except import ordering; no other pinned blob touched. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01GNFwcRAZnR3NXKahhz4DZe
random.Random(seed) drives permutation nulls and simulations, never secrets; annotate with the repo's existing '# nosec <id> # reason' form. Bandit over core/ contracts/ neosynaptex.py: 3 issues -> 0. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01GNFwcRAZnR3NXKahhz4DZe
tests/dopa_closure_v2 H4/H5 suites import torch at module level (the H4 file is evidence-pinned and cannot gain an importorskip), so Benchmarks failed at collection and Verify/Coverage would too. Install torch==2.10.0 (the torch_version recorded in evidence) from the CPU index in Verify, Coverage and Benchmark Regression. actionlint 1.7.12: all workflows clean. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01GNFwcRAZnR3NXKahhz4DZe
b4d9a36 labelled construct_gate with scope ROUTING_ONLY and evidence_level UNATTESTED_CALLER_DECLARATION in construct_gate.json and in inadmissible results, but the admissible final verdict still emitted bare asdict(verdict), so a successful run's final_verdict dropped the 'caller declaration only' marker. test_claim_manifest_hash_is_bound_to_ ordered_inputs (added by the same commit) was red on main; now green. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01GNFwcRAZnR3NXKahhz4DZe
The reproducibility test asserted bitwise equality of every float against results/m_invariant_scale_diagnostics/run_v1.json; on GitHub runners (different NumPy/BLAS build) values drift by 1 ulp (e.g. H 2.206603549280792 vs 2.2066035492807923), so Benchmark Regression could never pass. Structure, keys, order, strings, ints and bools remain exact; floats rel 1e-12. A 1e-9 relative perturbation of the committed evidence still turns the test red. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01GNFwcRAZnR3NXKahhz4DZe
On PR #20 the serial suite exceeded its job limits: Verify (3.12) cancelled at 25 min, Coverage at 35 min, Benchmark Regression at 20 min (3.10/3.11 finished at ~17 min). Slowest tests are pre-existing gamma/falsification reproductions (up to ~3.5 min each under load). pytest-xdist 3.8.0 -n auto in Verify, Coverage and Benchmarks; locally -n 10: 2674 passed, 29 skipped, 0 failed in 6m09s. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01GNFwcRAZnR3NXKahhz4DZe
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.
claim_status: measured
Purpose
Bring
mainto a verified green canonical state. The base commit04f232e2was never verified: the 14 PRs #5–#18 were merged on 2026-10-16 while GitHub Actions were blocked by billing, so no job ran. Every push workflow on main has now been rerun; this PR fixes what that rerun exposed.Changes
dopa_closure_v2.py).torch==2.10.0, the version recorded in evidence, is installed for the full-suite jobs. Without it the H4/H5 tests failed at collection.evidence_level: UNATTESTED_CALLER_DECLARATION. The test for it, added in b4d9a36, was red on main.Verification
🤖 Generated with Claude Code
https://claude.ai/code/session_01GNFwcRAZnR3NXKahhz4DZe