You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
http-capability-gateway tier-2 production-wiring.ADR-0004 is Accepted but not production-wired: Trustfile.a2ml tier-2 = PENDING; stated gaps = mTLS-as-primary path, E2E verification, benchmark evidence (~8–12 wks per the ADR). This is the real blocker for a governed public surface — until it lands the unified core stays internal/loopback and consumer K9 gates stay red (tracked, not masked). Tracks the existing ADR-0004 + audit; does NOT duplicate them.
Channel progress: Phase A (#96) opened as boj-server#78 — A1 contract / A2 authoring workflow / A3 example policy, specification-only. Sequential order intact; B–E (#97–#100) not started. No issue closed (joint-close).
Re-verified 2026-08-27 — the blocker holds, the gap list does not
The blocker is real and unchanged:boj-server/.machine_readable/contractiles/trust/Trustfile.a2ml line 900 still reads status: "PENDING — http-capability-gateway wiring forthcoming".
But all three "stated gaps" now have in-tree artifacts
gap
artifacts found
mTLS
src/abi/MTLSPolicy.idr, test/mtls_test.exs, docs/mtls-rotation-runbook.adoc, 7 test fixtures
⚠️Presence is not passing. Nobody has verified these go green — and given the estate-wide startup_failure epidemic (#657), their CI may never have run at all. That is the actual next step, and it is cheap.
The "~8–12 weeks" scope no longer describes the work. Recommend: run the suites, then either flip line 900 or re-file against whatever actually fails.
Activity
hyperpolymath commentedon May 18, 2026
Channel progress: Phase A (#96) opened as boj-server#78 — A1 contract / A2 authoring workflow / A3 example policy, specification-only. Sequential order intact; B–E (#97–#100) not started. No issue closed (joint-close).
Phase A — http-capability-gateway↔BoJ contract + policy authoring wor…
hyperpolymath commentedon May 18, 2026
Channel: Phase A (#96) merged — boj-server#78. Contract §3 records the load-bearing finding (router loopback trust-bypass ⇒ mandatory
X-Trust-Levelstrip + back-side isolation) which feeds Phase B/C. Sequential order intact; #97–#100 not started. No issue closed (joint-close).Phase C: E2E verification + gateway↔BoJ seam tests
Phase C: E2E verification + gateway↔BoJ seam tests (#11)
Phase C: declare gateway↔BoJ-gnosis seam in Trustfile [SEAMS]
Phase C: declare gateway↔BoJ-gnosis seam in Trustfile [SEAMS] (#90)
24 remaining items
feat(scripts): hcg-policy-smoke.sh — unknown-path default-deny canary…
feat(scripts): hcg-policy-smoke.sh — unknown-path default-deny canary…
feat(scripts): hcg-policy-smoke.sh — stealth-profile status canary (P…
docs(integration): Phase E observability spec (standards#100)
docs(integration): Phase E observability spec (standards#100) (#226)
feat(hcg): add surface-drift check script (standards#100 §1.5)
feat(hcg): add surface-drift check script (standards#100 §1.5) (#228)
ci(hcg): wire surface-drift check into Actions (standards#100 §1.5) (#…
feat(hcg): add spec-coverage check script (standards#100 §1.5) (#230)
docs(integration): Phase E §1.4 prereq — BoJ-side observability spec …
hyperpolymath commentedon Aug 27, 2026
Re-verified 2026-08-27 — the blocker holds, the gap list does not
The blocker is real and unchanged:
boj-server/.machine_readable/contractiles/trust/Trustfile.a2mlline 900 still readsstatus: "PENDING — http-capability-gateway wiring forthcoming".But all three "stated gaps" now have in-tree artifacts
src/abi/MTLSPolicy.idr,test/mtls_test.exs,docs/mtls-rotation-runbook.adoc, 7 test fixturese2e_test.exs,e2e_boj_integration_test.exs,e2e_property_test.exs,forged_trust_e2e_test.exsbench/baseline.json,gateway_latency.exs,docs/perf-contract.adoc,docs/RELEASE-CRITERIA.adocstartup_failureepidemic (#657), their CI may never have run at all. That is the actual next step, and it is cheap.The "~8–12 weeks" scope no longer describes the work. Recommend: run the suites, then either flip line 900 or re-file against whatever actually fails.