release/consolidation-2026-06-26: weft-seam conformance + project-root artifacts by tachyon-beep · Pull Request #69 · foundryside-dev/wardline

tachyon-beep · 2026-06-26T05:30:52Z

Consolidation release of two bodies of work onto main.

1. Weft-seam conformance program (this cycle)

Brings every weft peer-to-peer interface to the SEI bar — a frozen contract + golden-vector oracle + fail-closed CI gate + two-layer drift alarm — enforced by a single lie-detector gate.

Enforceable seam registry + gate (tests/conformance/test_seam_registry.py, seam_registry.json): a fail-closed, unmarked-default-suite gate that parses the 3 real marker sources; at_bar requires a real oracle + byte-pin + applied marker. Verdicts: at_bar 12, peer_conformant 4, gap 20, deferred 1, one_sided_na 3.
Shared WeftHttp transport + federation-status envelope single-sourcing (dedup across MCP/CLI/scan-jobs), behaviour-preserving.
12 at_bar seams incl. G1 (the wardline→legis signed scan-artifact wire) reaching the bar via a single byte-identical shared vector — wardline's real signer reproduces legis's byte-exact expected_signature, byte-pinned to legis's copy.
Legis fence lifted: the legis-involving seams were brought to the bar in their own repos (legis/loomweave/filigree) — SEI(loomweave→legis), git-rename, governed sign-off(→filigree), loomweave HMAC. Where wardline isn't a party, rows carry the new peer_conformant verdict (gated to require peer repo + commit + test evidence, never bare gap). Genuinely-incomplete seams (plainweave consumer unbuilt; two warpline obligations) stay gap with structured notes — no manufactured contracts.

2. Project-root-anchored artifacts + doctor hygiene (from `…06-25`)

Default scan artifacts anchor to the weft-project root; wardline doctor --repair gitignores the artifacts dir and sweeps stray managed artifacts (CLI + MCP doctor with destructiveHint); the wlfp2 move-stable fingerprint rekey (with fan-out). Already reviewed on the feat/project-root-anchored-artifacts line.

Merge integrity

The weft feature was merged clean (the 3 superseded artifacts-spec drafts dropped). All 5 conflicts were resolved by combining both sides (WeftHttp + URL redaction; federation_status imports ∪ redact; CHANGELOG [Unreleased] + [1.0.7]; 37 glossary line-anchors recomputed against merged source — verified no doc content lost). 3 frozen goldens re-frozen, each root-caused as a legitimate …06-25 feature and guard-verified to change only the expected fields.

Verification

Full suite 4419 passed, ruff clean, wardline scan --fail-on ERROR exit 0. Version: 1.0.7.

🤖 Generated with Claude Code

This ensures that the output of these potentially unsafe deserialization functions correctly sheds validation provenance (becomes `UNKNOWN_RAW`) in the taint propagation logic. Co-authored-by: tachyon-beep <544926+tachyon-beep@users.noreply.github.com>

Bumps the actions-minor-patch group with 1 update: [pypa/gh-action-pypi-publish](https://github.com/pypa/gh-action-pypi-publish). Updates `pypa/gh-action-pypi-publish` from 1.13.0 to 1.14.0 - [Release notes](https://github.com/pypa/gh-action-pypi-publish/releases) - [Commits](pypa/gh-action-pypi-publish@ed0c539...cef2210) --- updated-dependencies: - dependency-name: pypa/gh-action-pypi-publish dependency-version: 1.14.0 dependency-type: direct:production update-type: version-update:semver-minor dependency-group: actions-minor-patch ... Signed-off-by: dependabot[bot] <support@github.com>

Bumps [actions/configure-pages](https://github.com/actions/configure-pages) from 5 to 6. - [Release notes](https://github.com/actions/configure-pages/releases) - [Commits](actions/configure-pages@v5...v6) --- updated-dependencies: - dependency-name: actions/configure-pages dependency-version: '6' dependency-type: direct:production update-type: version-update:semver-major ... Signed-off-by: dependabot[bot] <support@github.com>

Bumps [actions/deploy-pages](https://github.com/actions/deploy-pages) from 4 to 5. - [Release notes](https://github.com/actions/deploy-pages/releases) - [Commits](actions/deploy-pages@v4...v5) --- updated-dependencies: - dependency-name: actions/deploy-pages dependency-version: '5' dependency-type: direct:production update-type: version-update:semver-major ... Signed-off-by: dependabot[bot] <support@github.com>

Bumps [actions/checkout](https://github.com/actions/checkout) from 4 to 7. - [Release notes](https://github.com/actions/checkout/releases) - [Commits](actions/checkout@v4...v7) --- updated-dependencies: - dependency-name: actions/checkout dependency-version: '7' dependency-type: direct:production update-type: version-update:semver-major ... Signed-off-by: dependabot[bot] <support@github.com>

Bumps [astral-sh/setup-uv](https://github.com/astral-sh/setup-uv) from 7.6.0 to 8.2.0. - [Release notes](https://github.com/astral-sh/setup-uv/releases) - [Commits](astral-sh/setup-uv@37802ad...fac544c) --- updated-dependencies: - dependency-name: astral-sh/setup-uv dependency-version: 8.2.0 dependency-type: direct:production update-type: version-update:semver-major ... Signed-off-by: dependabot[bot] <support@github.com>

The waiver_add entity_symbol path resolves a qualname through Loomweave (SeiResolver.detect + resolve_locator) before writing the waiver — an outbound/loopback network side effect. But the tool declared only READ|WRITE and _effective_tool_capabilities had no waiver_add branch, so ToolPolicy never denied it under allow_network=false: any client allowed the write tools could trigger signed Loomweave requests despite the network being fenced off. Add a waiver_add branch that declares NETWORK under the exact predicate that fires the resolve — entity_symbol present, entity_id absent (entity_id wins and is carried opaque), and a Loomweave URL configured — mirroring the scan/explain_taint/dossier gates. Also align the handler so it only builds the Loomweave client when entity_symbol and not entity_id, matching the declared side effect. resolve_entity_binding_input is the sole network path in _waiver_add (add_waiver is a local FS write); the gate is a precise, fail-closed match. Regression tests: entity_symbol under no-network policy is denied before the handler runs; entity_id-only and entity_id-wins-over-symbol stay ungated. Closes wardline-14359d070b. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

The Level-2 branch-join merges deduplicated candidates with a nested linear scan of the growing candidate list: `any(lam is seen for seen in bucket)` in _merge_branch_bindings and `if fqn not in bucket` in _merge_branch_types. That is O(bucket) per insert, O(bucket**2) per merge. Across a chain of N one-armed branches rebinding the same name (`if flagK: cb = lambda c: sinkK(c)`), the candidate set grows to N over N merges -> O(N**3). An attacker-authored file with ~1100 such branches drove a DEFAULT-gate scan to ~15s, exhausting CPU on every local and CI run (wardline-c797baf28b). Both merges now dedup via a per-name identity/equality set: O(1) per insert, O(bucket) per merge, O(N**2) cumulative. The change is behavior-identical -- same candidate set, same first-seen insertion order, same dedup semantics -- so no false negative is introduced. A cap was deliberately avoided: dropping candidates would be an FN, against this module's no-FN discipline; this mirrors the d7ecb90 precedent of eliminating the complexity soundly rather than capping. The 1100-branch PoC drops 4.388s -> 0.080s (55x); the var_types sibling 1.015s -> 0.037s (27x). Full scanner suite (1363 tests) stays green. Tests: - test_lambda_candidate_merge_is_not_cubic_on_chained_rebinds (DoS guard) - test_var_type_candidate_merge_is_not_cubic_on_chained_rebinds (sibling) - test_chained_one_armed_rebinds_keep_every_lambda_candidate (no-FN lock) Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

Passed `_SAFE_GIT_CONFIG = ("-c", "core.fsmonitor=false")` to all untrusted Git subprocess executions in delta and legis. Co-authored-by: tachyon-beep <544926+tachyon-beep@users.noreply.github.com>

… consumer drift check Phase C of warpline integration. Publishes the versioned, drift-checked wardline.delta_scope.v1 producer artifact (the DeltaScopeReport.to_dict() shape, 13 fields incl. scope_source + producer_generated_at) with a drift test, and extends the warpline.reverify_worklist.v1 consumer conformance test to assert generated_at capture plus a WARPLINE_REPO-gated published-artifact drift marker (skips clean until warpline publishes its half). Closes the contract-integrity portion of wardline-c0563eee74. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

…t whole-file Final-review Critical: the attest-2 boundary content_hash is EntityBinding.content_hash = Loomweave identity-resolve content_hash, which per the Accepted ADR 2026-06-02-wardline-hash-granularity-two-model is the ENTITY-BODY span hash (same granularity as Filigree content_hash_at_attach), NOT whole-file. The published artifacts had it inverted (whole-file / 'not entity-span' / 'do not key as entity-precise') — a fail-open instruction that contradicted a live conformance test and would make warpline cross-compare granularities (permanent false-STALE) or transfer a clean verdict across sibling entities. The data was always correct (entity-body is the right key for a SEI-scoped boundary); only the labels were wrong. Relabelled across contract doc, attest.py comment, MCP _ATTEST_OUTPUT_SCHEMA (+ golden re-frozen, VENDORED_BLOB_SHA cd08480), attestation.md, CHANGELOG, seam_registry. Also (Minor): noted verdict=='defect' as a proven-bad signal distinct from the non-exhaustive enrichment_reasons triple; marked the WARPLINE_REPO-gated test as an existence-only placeholder. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

…ness (D1) Replaces the phantom producer_generated_at (which read data.generated_at — a key warpline never emits) with producer_completeness: a VERBATIM, unverified echo of warpline's now-shipped data.impact_completeness object (self-assessed completeness + staleness of the change-impact analysis, in one object). Captured defensively (dict-or-None), namespaced, and never fed into mode/gate_authority/any verdict — wardline stays the trust authority. Threaded AffectedScope -> run_scan -> DeltaScopeReport.to_dict() (auto-propagates CLI/SARIF); MCP scope schema mirrors it as a loose object (wardline echoes verbatim, does not pin warpline's inner shape); key-parity test + wardline.delta_scope.v1 fields updated; golden re-frozen. Realigned the vendored warpline_delta fixtures to warpline's REAL published envelope (full reverify_worklist.v1 shape: completeness/impact_completeness/staleness/items/ resolved/unresolved) and made the WARPLINE_REPO-gated drift test REAL — it now validates every fixture against warpline's published contracts/reverify_worklist.v1.schema.json via jsonschema (was existence-only). Verified live: WARPLINE_REPO=/home/john/warpline -> 10/10 pass. Closes D1; the warpline<->wardline reverify seam is now two-sided drift-checked (wardline-c0563eee74). Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

…ung-2) Agent-facing operational counterpart to the wardline-attest-2 contract: the prompt warpline's risk-as-verification step uses to turn a bundle into proven_good / proven_bad / risk=unavailable, with wardline as the sole trust authority. Aligned to the shipped contract: content_hash is ENTITY-BODY span (entity-precise; compare same-granularity only, never whole-file), defect is a distinct proven_bad signal, and the enrichment_reasons triple is non-exhaustive (+ content_drift). Cross-linked from wardline-attest-2.md. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

tachyon-beep · 2026-06-27T01:53:10Z

Added: warpline integration (item 4, P1) — `9f5924bf..b1a97d5e`

Six commits implementing the wardline↔warpline P1 integration (evaluated + planned earlier this cycle):

(a) delta-scope provenance — scope_source + warpline's UNVERIFIED producer_generated_at in the --affected scope block (9f5924bf); published+drift-checked wardline.delta_scope.v1 (ab0754fe).
(b) attest-2 — bumped attest bundle to wardline-attest-2 with per-boundary content_hash binding (dffa092d); final-review Critical fix: content_hash is entity-body span (= Loomweave EntityBinding/Filigree content_hash_at_attach granularity per ADR 2026-06-02-hash-granularity-two-model), not whole-file — relabelled across all published artifacts (7b4afcda).
(D1) warpline-published completeness mirrored as producer_completeness (33d5d5f2).
Consumer prompt — agent-facing docs/contracts/wardline-attest-2-consumer-prompt.md for warpline's Rung-2 risk-as-verification, aligned to the corrected contract (b1a97d5e).

Boundary preserved: wardline never calls warpline (consumes pushed/untrusted scope); warpline never declares clean (relays wardline's fail-closed 3-valued verdict; unknown/drift → risk=unavailable).

Gate (local, full tree): pytest 4436 passed / 1 skipped · ruff check clean · mypy clean (530 files).

⚠️ Pre-existing (NOT warpline-p1): ruff format --check flags ~12 files (tests/unit/rust/*, etc.) already present at df0a79f8 — outside this work's scope; flagging for the release owner to address before merge if CI gates on format.

Pure formatting (no logic change) across 12 files flagged by `ruff format --check` on the release branch — whitespace/line-wrapping in tests + install/doctor.py. Brings the branch's Lint+Format CI check green so PR #69 is merge-ready. No warpline-p1 source files affected. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

A sibling's .weft/<sibling>/ephemeral.port advertises a live `serve` instance. When that process exits or wedges, the file lingers and every `wardline scan` dials a dead/hung origin, stalling the agent gate up to the federation 30s urlopen timeout per round-trip on a purely advisory emission — the reported ~90s `wardline scan .` hang in dogfood repos (~/elspeth). New `stale_sibling_ports` doctor check (CLI + MCP `doctor` tool): - Probes each advertised port (filigree, loomweave) at the host the scan dials — filigree→localhost (self-heals over IPv4/IPv6 so a ::1-only filigree is not falsely cleared), loomweave→127.0.0.1 — with a short 2s deadline, NOT the 30s federation default. - Unreachable (connection refused OR no HTTP reply within the deadline — a wedged server) is stale; --repair / fix:true deletes the file so resolve_*_url falls back to "no sibling" and the scan stops dialing it. - A live server (any HTTP status) is never touched; the delete is regular-file / no-follow confined (a symlinked ephemeral.port is never followed). Advisory like the stray-artifact sweep — never flips the aggregate doctor verdict. Verified end-to-end against ~/elspeth: a wedged filigree port that made `wardline scan .` take 1m47s clears in ~4s, and the subsequent scan completes in 24s (pure CPU) with the wedged trap still up. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

Make durable the bet rotation a prior 2026-06-27 session decided but crashed before checkpointing. PDR-0002 records the within-grant Later→Now promotion (Codex hardening close-out paid off — both P1s closed, codex-security-2026-06-20 batch 0 open, G2 at target → ACCEPT). Commits the orphaned metrics.md (G2 + G2-seam readings) and PRD-0002; rotates roadmap.md Now; rewrites current-state.md; re-stamps the grant review date (re-confirmed unchanged). Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

…'t read my store" Extend the MCP doctor tool with a READ-ONLY repo-binding probe so a stale-but-running wardline (server starts, --root resolves, but it cannot read its repo-scoped baseline store) reports honestly instead of looking healthy — the wardline analog of the 2026-06-26 loomweave silent-de-attach incident, and the producer half of lacuna's MCP-attachment regression harness (weft-seam-conformance Now bet, PDR-0002 / G2-seam). - inspect_baseline_store(root) -> BaselineStoreStatus (core/baseline.py, next to BASELINE_VERSION; reuses require_yaml + _build_baseline; never writes/migrates/mkdirs). The non-tautological signal is the schema version READ STRICTLY from inside the store — null unless an int version was actually read (a degenerate empty store -> null + binding_ok false, never the served constant), so "root == cwd" can never fake a pass. - _check_repo_binding(root) (install/doctor.py) emits a top-level repo_binding block {resolved_root, store{present,readable,schema_version,baseline_finding_count}, binding_ok} + a uniform doctor.repo_binding check. Fork-1 split (honors the not-noisy anti-goal): present-but-unreadable (version mismatch/corrupt — the incident) is status=error and flips doctor.ok; ABSENT (baseline is opt-in) stays status=ok and never nags a baseline-less repo. - repo_binding declared in _DOCTOR_OUTPUT_SCHEMA so the structuredContent emission carries it; golden re-frozen (mcp_output_schemas.golden.json + VENDORED_BLOB_SHA) per the re-freeze procedure. - Trust-boundary: the unreadable diagnostic is content-free (names only the store filename + served version) — a crafted store can no longer echo its content back through the seam. Consumer contract (lacuna probe reads verbatim): repo_binding.binding_ok and repo_binding.store.schema_version; predicate binding_ok==true AND schema_version not null. Full suite green (4472 passed); ruff/mypy clean; wardline self-gate exit 0 / 0 active. Built via ultracode workflow (1 TDD implementer + 5 adversarial verifiers + gate), all lenses pass. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

…-0003 Record this session's delivery on the weft-seam-conformance Now bet: the wardline doctor.repo_binding store-read seam (commit c661286). PDR-0003 captures the load-bearing call — the Fork-1 split (absent baseline stays silent, only an unreadable/corrupt store errors) deviating from the literal lacuna prompt to honor the not-noisy anti-goal, contract-safe by construction and live-proven. Adds a dated G2-seam reading (new honesty surface landed + round-trip-proven; 6-set target unchanged); refreshes current-state.md. Roadmap untouched (no horizon change). Nothing escalated — the editable global-install reinstall was owner-confirmed in-session. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

Formal ACCEPT of the Codex hardening bet, deferred across two sessions. Judged against all 5 PRD-0001 criteria with evidence re-run at HEAD: c797 DoS bound pinned O(N^2) (test_lambda_candidate_merge_is_not_cubic_on_chained_rebinds), d96b credential gate fail-closed (test_check_does_not_send_token_to_project_published_port), G1 precision held via the no-candidate-dropped soundness-lock family + full suite 4472 + dogfood 0-active. All met -> ACCEPT. PRD-0001 status -> ACCEPTED; dated G2 reading added; current-state ACCEPT item retired. Criterion-3 honesty note recorded (verified the no-finding-dropped property via durable soundness-lock tests, not a one-time pre/post byte-diff that would need risky pre-fix checkout on the shared tree). Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

…(crit-3b) Make the SEI-oracle and warpline reverify-worklist producer-SOURCE drift checks required & fail-closed in CI (PRD-0002 crit-3b; wardline-79ba05f464 / wardline-c0563eee74). TAXONOMY CHANGE (src semantic, not just a CI job): sei_drift + worklist_drift are added to wardline._live_oracle.LIVE_ORACLE_MARKERS, so an armed WARDLINE_LIVE_ORACLE_REQUIRED=1 run turns a missing-source SKIP into a FAILURE. Until now _drift markers were the skip-clean release-gate tier; they now fail closed because the new weekly source-drift job provisions their sibling source. - .github/workflows/ci.yml: new `source-drift` job (schedule/dispatch) checks out loomweave + warpline origin/main (read-only WARDLINE_SIBLING_SOURCE_TOKEN) and runs `pytest -m "sei_drift or worklist_drift"` fail-closed. - src/wardline/_live_oracle.py: add the two markers to LIVE_ORACLE_MARKERS. - tests/unit/test_live_oracle.py: lock the contract (the two are in; the four unrun _drift markers stay out, fail-closed-only-when-armed). - tests/conformance/test_warpline_delta_scope.py: normalize the published-schema check to the standard WARDLINE_WARPLINE_REPO (was a dead WARPLINE_REPO env var); kept OUT of the fail-closed job until warpline pushes its schema to origin/main. - tests/conformance/test_seam_registry.py: update the marker-taxonomy doc block. - docs/plans/2026-06-28-weft-seam-crit3-ci-drift.md: reality-validation + plan. crit-3a (wardline.delta_scope.v1 producer artifact + its in-CI drift check) was already shipped; this closes the consumer/producer source-drift CI leg. Full suite 4475 passed; ruff/mypy clean; vendored fixtures verified byte-synced to loomweave + warpline origin/main. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

…3.12==3.13) Python 3.13's ast.dump (new default show_empty=False) OMITS empty-list fields (posonlyargs=[], decorator_list=[], type_params=[], ...) that <=3.12 emit. The entity-body discriminator hashed ast.dump output, so wardline under 3.12 vs 3.13 minted DIFFERENT fingerprints for the same source — and the fingerprint is the cross-tool JOIN KEY (baseline/waiver/judged stores + the Filigree wire). The identity parity corpus (frozen on 3.13) drifted on 3.12; 5 golden tests failed on the 3.12 CI matrix leg (test_identity_corpus_is_byte_identical x3, test_golden_matches_live_producer, test_builtin_findings_match_golden), all from this one root cause. Fix: _canonical_ast_dump reproduces the 3.13 show_empty=False canonical form on EVERY interpreter (verified node-for-node equal to 3.13's ast.dump across the fixtures), so the fingerprint is byte-identical 3.12==3.13. Done STRUCTURALLY, not by regex-stripping the dump string, because a string literal like "x=[]" renders as Constant(value='x=[]') and a text strip would corrupt the join key. 3.13 output is byte-unchanged (cdump==ast.dump there), so NO corpus re-freeze and NO fingerprint-scheme bump — only the broken 3.12 values converge to the 3.13 reference. - src/wardline/scanner/rules/_fingerprint.py: structural _canonical_ast_dump. - tests/unit/scanner/rules/test_entity_fingerprint_stability.py: pin the canonical form + the string-literal-preservation guard. Verified: full suite 3.12 4478 passed (was 5 failed), 3.13 unchanged; the 5 goldens green on both interpreters; ruff/mypy clean. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

…oss-interpreter determinism fix; PDR-0005, PDR-0006 Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

Closes wardline-a0eaa7dd12 (and subsumes wardline-7971cbcf9e). The arch-analysis H1 finding framed ~102 deferred imports as "masking real cycles." Graph analysis showed the runtime module graph was already acyclic and only TWO latent runtime cycles remained, each closed by a tiny set of deferred edges; the other ~99 deferred imports are legitimate lazy loads (optional federation, CLI/MCP fast-start, narrow import surface). So the fix is cycle-elimination + enforced layering, not chasing the count to zero. Cycles broken (combined runtime graph now fully acyclic): - scanner.grammar <-> scanner.rules: extracted the boundary primitives (BoundaryType/LevelArg/BUILTIN_BOUNDARY_TYPES) into engine-floor scanner/boundary_types.py; grammar re-exports them as a facade; the 3 cycle members + doctor import the primitive directly. - run -> suppression -> finding_identity -> baseline -> run: relocated the scan-running orchestration (collect_and_write_baseline, generate_baseline) out of policy-tier core/baseline.py into surface-tier core/baseline_ops.py; hoisted two redundant deferred gate_breakdown imports in run.py. Deferred-import count 103 -> 92 (cycle fix + 8 redundant promotions where the target was already module-level imported); cycle-MASKING count driven to 0. Enforcement (the durable guard): - 2 enforcing import-linter contracts (engine-purity, policy-purity); CI `lint-imports || true` -> gating; stale "BROKEN" comment removed; added lint-imports to `make lint`. - tests/conformance/test_import_layering.py: combined-graph acyclicity (the intra-tier guard contracts can't express) + a Python tier-purity backstop over EVERY module + a contract-module-resolve drift guard + absolute-import check. make ci green (4486 tests, 92.40% cov), lint-imports 2 kept/0 broken. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

…ibility wardline fires only when untrusted data crosses a DECLARED trust boundary (@trusted / @external_boundary / @trust_boundary, builtin or pack). A codebase that declares none produces zero defects no matter what it does, so a `wardline scan . --fail-on ERROR` gate over it passes green while checking nothing. The elspeth FastAPI gate hit exactly this; the only prior hint was buried INFO-severity WLN-L3-LOW-RESOLUTION — the severity agents filter out. Adds a scan-level enforcement-posture verdict derived from the engine's existing WLN-ENGINE-METRICS finding (taint_source_counts.anchored/config = recognized boundaries; histogram sum = functions analyzed). A scan that recognized ZERO boundaries over a non-trivial codebase is flagged INERT — no engine change: - core/resolution_posture.py: compute_resolution_posture() (pure read of findings) - cli/scan.py: Python inert-gate stderr banner — the counterpart of the existing Rust empty-trust-surface anti-false-green warning - core/agent_summary.py: structured `resolution` block (the agent-facing surface) - mcp/server.py: `resolution` in the scan outputSchema (+ re-froze the MCP schema golden) Calibrated live: elspeth -> inert=True; wardline corpus (anchored=43) -> inert=False. Part A of wardline-bd9d1e65cb. blake3 was a red herring; FastAPI source coverage is moot without declared boundaries (pack-bridge tracked separately). Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

The elspeth gate is inert because wardline is annotation-driven and doesn't read elspeth's OWN trust vocabulary — elspeth annotates ~25 external-data boundaries with elspeth.contracts.trust_boundary.trust_boundary(tier=3, source_param=...), which wardline ignores, so it recognizes zero boundaries (blake3 and FastAPI source coverage are both irrelevant to this — proven in wardline-bd9d1e65cb). Rather than re-annotate elspeth in wardline's vocabulary, bind elspeth's existing vocabulary with a pack: a BoundaryType mapping the decorator to wardline's validating-boundary seed (EXTERNAL_RAW args -> ASSURED return). elspeth's tier/source/source_param kwargs are not trust LEVELS, so level_args=() and the matcher ignores them. Validated end-to-end on an elspeth-shaped target: - WITH the pack: both @trust_boundary functions recognized (scan non-inert, recognized_boundaries=2); a boundary that returns its untrusted source_param unvalidated fires PY-WL-119 ERROR. - WITHOUT the pack: zero boundaries recognized, no defect (the inert state). The pack (tests/grammar/fixtures/elspeth_trust_boundary_pack.py) is the copy-ready deliverable for elspeth: place on the import path + reference under [wardline] packs in weft.toml (or --trust-pack). Full elspeth-repo FP calibration over the 25 real boundaries is the remaining follow-on (tracked). Part of wardline-bd9d1e65cb. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

…ridge; PDR-0007, PDR-0008 Owner-directed investigation of the ~/elspeth/ report. Corrected diagnosis (wardline is annotation-driven; blake3 a red herring), shipped Part A inert-gate visibility (b3d0a81) and the elspeth-vocab pack-bridge (72bc9eb). Now bet (weft-seam-conformance) unchanged. Flags: PR #69 release scope grew; install pack in elspeth; auto-inference strategic question (vision tension) escalated; pack FP UNMEASURED (calibration follow-on).

…rust vocab elspeth is an unrelated project; no elspeth-specific code belongs in wardline. The committed pack-bridge fixtures embedded `elspeth.contracts.trust_boundary` into the suite. Genericize to a fictional `acme.security.trust_boundary` vocab so the test still proves the GENERIC pack-bridge mechanism (a third-party @trust_boundary decorator whose kwargs are not trust levels maps to a validating-boundary seed; PY-WL-119 fires WITH the pack, inert WITHOUT) without naming an unrelated project. - tests/grammar/fixtures/{thirdparty_trust_boundary_pack,target_thirdparty_boundary}.py + tests/grammar/test_thirdparty_pack_bridge.py replace the three elspeth_* files. - Scrub the two incidental elspeth mentions in src/ comments (resolution_posture.py, judge.py) and the Part A test names/comments (test_resolution_posture, test_cli, test_scan_inert_posture) to generic "framework app" phrasing. `grep -rI elspeth src/ tests/ docs/reference --include='*.py'` is now empty. Part of wardline-bd9d1e65cb. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

… flag Part B of wardline-bd9d1e65cb. Two new doctor checks, wired into both the machine-readable surface (machine_readable_doctor → MCP `doctor` / `--fix`) and the CLI text path. - engine.selftest: runs the taint analyzer on a tiny BUILT-IN @trusted source→sink fixture (flask request multidict → os.system) in a throwaway tempdir and asserts PY-WL-108 fires → "taint analysis fires correctly". Proves the analysis pipeline is wired and FIRING in this install. Deliberately worded engine-not-target: a passing self-test does NOT mean the user's scans enforce (an annotation-free codebase is still inert by design — Part A's per-scan posture carries that). The fixture lives outside the source tree so it can never enter wardline's own corpus golden. - loomweave.dep: flags a configured-but-uninstalled Loomweave integration. VERIFICATION FINDING (the issue flagged "verify the config shape first"): there is NO [wardline.loomweave].url config key — resolve_loomweave_url reads only the --loomweave-url flag / WARDLINE_LOOMWEAVE_URL env / published port. "Configured" = explicit operator intent (launch flag, env, or .mcp.json --loomweave-url arg); ambient published-port auto-discovery is deliberately NOT flagged (the write is fail-soft, and flagging it would nag every base install near a Loomweave sibling). require_blake3() raises LoomweaveError when the [loomweave] extra is absent. _mcp_filigree_url generalized to _mcp_wardline_arg(root, flag) with a _mcp_loomweave_url sibling. MCP output golden unchanged (it is a schema; checks is a generic array). Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

Part C of wardline-bd9d1e65cb. wardline shipped Flask request coverage but none for FastAPI/starlette, so a FastAPI+pydantic app's request entry points were invisible and a `wardline scan --fail-on ERROR` gate over it was inert. Annotation-based source seeding via the typed-receiver + curated-member model (NOT whole-param tainting): a parameter whose annotation resolves (via the alias map, never the parameter NAME) to fastapi.Request / starlette.requests.Request is a request boundary, and curated DATA members read off it yield EXTERNAL_RAW — .query_params/.path_params/.headers/.cookies (property reads) and .json()/.body()/ .form()/.stream() (method calls). Framework objects (.app/.state/.url/.scope/.client/ .base_url) stay clean; a bare uncalled `req.json` (a coroutine-method object) stays clean. The mechanism is general (a request-type → member table); seeding is scoped to the two FQNs deliberately (precision over recall). - variable_level.py: `_REQUEST_SOURCE_TYPES` table + two seed points (Seed A = property read in _resolve_expr's Attribute handler; Seed B = method call in _resolve_call's typed-receiver block, after the project-summary dispatch so a real summary wins first). - project_resolver.py: _RESOLVER_VERSION sp1e→sp1f so warm/persisted summaries of request-handler modules recompute rather than serve stale-CLEAN results. Return-pass FN closure (adversarial review): _CURRENT_VAR_TYPES is reset before compute_return_taint, so a @trusted boundary that DIRECTLY returns a request source (`return req.query_params.get('x')`) was summarized clean and its PY-WL-101 boundary-leak went undetected — while the local-var form already fired. Re-establish the receiver-type map for the return VALUE pass, FILTERED to request FQNs (so the general typed-receiver dispatch stays inert in return-taint — zero blast radius on `return typed.method()`), scoped to compute_return_taint (not the explain-only compute_return_callee). Corpus byte-identical (wardline's own src imports no fastapi/starlette; both seeds are structurally dead on it). Tests: 14 must-fire + 12 must-not-fire + 3 return-leak + 1 freedom-zone-return. Annotation-driven throughout — an undecorated handler fires nothing; this makes the SOURCE visible, it does not declare the boundary. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

…ck shift The Part A inert-gate commit (b3d0a81) added lines to cli/scan.py and core/ agent_summary.py, shifting the summary-block and agent-summary-key line numbers the finding-lifecycle glossary cites — leaving tests/docs/test_glossary_vocabulary.py (the two-way doc↔code anchor lock) red on the branch. Re-point both the `_ANCHORS` table and the glossary citations to the current lines (scan.py suppressed/active/gate → 569/570/622; agent_summary.py keys +1, informational-display 165→172). No behavior change; restores the branch to green. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

…-gate prevalence; PDR-0009 Owner decision A+C: do not make the framework-auto-inference vision change; instrument inert-gate prevalence (within grant). Baseline 2026-06-29: of 9 armed-gate repos, 5 framework-shaped, 5/5 inert, but realized reliance-gated harm = 1. Option B (per-parameter engine change + vision change) parked, gated on a metric-bound non-self-sealing trigger (>=5; baseline 1). Reconciled drift: Part B/C now DONE+committed, dogfood ticket bd9d1e65cb CLOSED. Grant re-confirmed 2026-06-29 (date-only). Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

Cut [Unreleased] -> [1.1.0]. Headline: FastAPI/Starlette request-source coverage, inert-gate visibility, third-party trust-vocab pack-bridge, doctor engine self-test + repo-binding/stale-port/repair, delta-scope. Security: cross-interpreter fingerprint determinism (3.12==3.13 join-key) + the agent-surface hardening batch. Owner-authorized release (v1.1.0 minor). Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

…-bump-robust CI red on the feature head: install/doctor.py (Part B) was never ruff-formatted, and test_package asserted version startswith('1.0.') which the 1.1.0 bump broke. Format the file; assert a 1.x semver shape (re.fullmatch r'1\.\d+\.\d+') so future point/minor releases don't re-break it. Local: ruff format/check clean, test passes. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

John Morrissey and others added 30 commits June 21, 2026 16:58

fix: stop repo gitignore hiding discovered source

ab09786

fix: prevent stdlib submodule spoof taints

e73124b

fix: reject MCP doctor caller filigree URLs

c0de30c

fix: harden rekey snapshot provenance

7d16b55

fix: preserve rust command shadow rebinds

b1a9de3

fix: drop untrusted MCP sibling URLs

0a30f62

fix: bound delta caller seed expansion

d7ecb90

fix: preserve explicit sibling loopback pins

09eae7a

docs: consolidate Wardline release materials

504cf87

docs: consolidate Wardline release materials

b5bdb7d

Fix git core.fsmonitor local execution vulnerabilities

073ec2f

Passed `_SAFE_GIT_CONFIG = ("-c", "core.fsmonitor=false")` to all untrusted Git subprocess executions in delta and legis. Co-authored-by: tachyon-beep <544926+tachyon-beep@users.noreply.github.com>

Fix git core.fsmonitor local execution vulnerabilities

8418c77

Passed `_SAFE_GIT_CONFIG = ("-c", "core.fsmonitor=false")` to all untrusted Git subprocess executions in delta and legis. Co-authored-by: tachyon-beep <544926+tachyon-beep@users.noreply.github.com>

fix(doctor): preserve filigree probe provenance

f6cf5e7

fix(filigree): soft-fail unsafe mint token reads

bc96534

fix(scanner): bound l2 taint analysis work

eff7c7c

fix(filigree): avoid response body amplification

5113b44

fix(site): pin site-kit fetch in pages build

42e2bab

fix(mcp): advertise scan integration effects

e27ed1b

fix(mcp): gate rekey cache-dir writes

de1509d

fix(rust): isolate mount overlay prepass failures

9747831

fix(fingerprint): bind singleton suppressions to source body

751af16

fix(rust): surface invalid trust markers

3e1db6b

John Morrissey and others added 4 commits June 26, 2026 23:56

John Morrissey and others added 24 commits June 27, 2026 11:55

fix(doctor): preserve probe provenance and gate stale probes

595e953

docs(plan): mark crit-3b T3 done + CI-verified (run 28301178826)

a1f121f

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

product: ACCEPT crit-3 (source-drift CI fail-closed) + fingerprint cr…

e4668ab

…oss-interpreter determinism fix; PDR-0005, PDR-0006 Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

docs: add 2026-06-28 architecture analysis

7e99147

fix: harden baseline and delta contracts

b5abe39

tachyon-beep merged commit 19be199 into main Jun 28, 2026
9 checks passed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

release/consolidation-2026-06-26: weft-seam conformance + project-root artifacts#69

release/consolidation-2026-06-26: weft-seam conformance + project-root artifacts#69
tachyon-beep merged 122 commits into
mainfrom
release/consolidation-2026-06-26

tachyon-beep commented Jun 26, 2026

Uh oh!

tachyon-beep commented Jun 27, 2026

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

Uh oh!

Conversation

tachyon-beep commented Jun 26, 2026

1. Weft-seam conformance program (this cycle)

2. Project-root-anchored artifacts + doctor hygiene (from …06-25)

Merge integrity

Verification

Uh oh!

tachyon-beep commented Jun 27, 2026

Added: warpline integration (item 4, P1) — 9f5924bf..b1a97d5e

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

2. Project-root-anchored artifacts + doctor hygiene (from `…06-25`)

Added: warpline integration (item 4, P1) — `9f5924bf..b1a97d5e`