Merge pull request #30 from dhawal-ss/codex/integrate-concurrent-hardening

Integrate concurrent production hardening

Author: dhawal-ss

.github/workflows/ci.yml

@@ -121,6 +121,10 @@ jobs:
     runs-on: ubuntu-latest
     steps:
       - uses: actions/checkout@11d5960a326750d5838078e36cf38b85af677262 # v4
+        with:
+          # The readiness validator binds sourceTreeHash to an exact historical
+          # sourceCommit, so a synthetic depth-one PR merge is insufficient.
+          fetch-depth: 0
 
       - name: Setup Node.js
         uses: actions/setup-node@49933ea5288caeca8642d1e84afbd3f7d6820020 # v4

.github/workflows/matrix-federation-acceptance.yml

@@ -82,20 +82,30 @@ jobs:
                   --unlock --components=secrets >/dev/null
                 npm run test:matrix-spike 2>&1 | tee "$MESH_ACCEPTANCE_LOG"
               '
+            docker compose -f infra/matrix-spike/docker-compose.yml \
+              down --volumes --remove-orphans 2>&1 |
+              tee "matrix-spike-logs/teardown-cycle-$cycle.log"
           done
 
           source_sha="$(git rev-parse HEAD)"
+          source_tree_hash="$(git rev-parse 'HEAD^{tree}')"
           cycle_1_sha="$(sha256sum matrix-spike-logs/acceptance-cycle-1.log | awk '{ print $1 }')"
           cycle_2_sha="$(sha256sum matrix-spike-logs/acceptance-cycle-2.log | awk '{ print $1 }')"
+          teardown_1_sha="$(sha256sum matrix-spike-logs/teardown-cycle-1.log | awk '{ print $1 }')"
+          teardown_2_sha="$(sha256sum matrix-spike-logs/teardown-cycle-2.log | awk '{ print $1 }')"
           printf '%s\n' \
             '{' \
             '  "schemaVersion": 1,' \
             "  \"sourceSha\": \"$source_sha\"," \
+            "  \"sourceTreeHash\": \"$source_tree_hash\"," \
+            "  \"workflowRunId\": \"${GITHUB_RUN_ID}\"," \
+            "  \"workflowRunAttempt\": \"${GITHUB_RUN_ATTEMPT}\"," \
+            "  \"workflowEvent\": \"${GITHUB_EVENT_NAME}\"," \
             '  "status": "passed",' \
             '  "independentResetTestCycles": 2,' \
             '  "cycles": [' \
-            "    { \"iteration\": 1, \"log\": \"acceptance-cycle-1.log\", \"sha256\": \"$cycle_1_sha\" }," \
-            "    { \"iteration\": 2, \"log\": \"acceptance-cycle-2.log\", \"sha256\": \"$cycle_2_sha\" }" \
+            "    { \"iteration\": 1, \"log\": \"acceptance-cycle-1.log\", \"sha256\": \"$cycle_1_sha\", \"teardownLog\": \"teardown-cycle-1.log\", \"teardownSha256\": \"$teardown_1_sha\" }," \
+            "    { \"iteration\": 2, \"log\": \"acceptance-cycle-2.log\", \"sha256\": \"$cycle_2_sha\", \"teardownLog\": \"teardown-cycle-2.log\", \"teardownSha256\": \"$teardown_2_sha\" }" \
             '  ]' \
             '}' > matrix-spike-logs/acceptance-report.json
 
@@ -118,7 +128,7 @@ jobs:
           name: matrix-federation-acceptance-${{ github.sha }}-${{ github.run_number }}
           path: mesh/matrix-spike-logs/
           if-no-files-found: error
-          retention-days: 14
+          retention-days: 30
 
       - name: Tear down disposable homeservers
         if: ${{ always() }}

.gitignore

@@ -99,6 +99,7 @@ claudereview/
 # Generated local integration-test state
 **/infra/matrix-spike/runtime/
 **/infra/homeserver/runtime/
+**/infra/homeserver/rotation-evidence/
 
 # Local configuration and secrets
 .env

MESH_PRODUCTION_READINESS_CONTINUATION_2026-08-01.md

@@ -1,5 +1,9 @@
 # Mesh production-readiness continuation
 
+> **ARCHIVED 2026-08-01 SNAPSHOT.** This is historical evidence, not a current
+> readiness assertion. Current source and the schema-v2 readiness ledger win;
+> rerun every named gate before reusing any count, SHA, or completion claim.
+
 Status: `ALL_LOCALLY_SPECIFIABLE_FINDINGS_COMPLETE_NOT_RELEASE_READY`
 
 Date: 2026-08-01

MESH_SOL_PRODUCTION_IMPLEMENTATION_PLAN_2026-07-31.md

@@ -1,5 +1,9 @@
 # Mesh — Competitive Audit & Production Implementation Plan — For ChatGPT Sol
 
+> **ARCHIVED 2026-07-31 PLAN.** Do not execute this file as a current handoff
+> and do not reuse its production assertions. Current source, `AGENTS.md`, and
+> the schema-v2 readiness ledger supersede it.
+
 **Date:** 2026-07-31
 **Branch:** `main`
 **Integrated implementation SHA:** `7effb0cea2eba0b92aa4a62d749aad12ddbfdbbe`

MESH_SOL_REFERENCE_REPOSITORY_PRODUCTION_PLAN_2026-07-31.md

@@ -1,5 +1,9 @@
 # Mesh reference-repository production implementation plan
 
+> **ARCHIVED 2026-07-31 REFERENCE PLAN.** It is retained for provenance and
+> design history only. Current source, `AGENTS.md`, and the schema-v2 readiness
+> ledger supersede every dated readiness, SHA, and implementation assertion.
+
 **Prepared for:** the next ChatGPT Sol implementation agent
 **Snapshot date:** 2026-07-31
 **Repository:** `D:\Creations\Applications\mesh`

PRODUCTION_BETA_PLAN.md

@@ -1,18 +1,23 @@
 # Mesh: Production Beta Implementation Plan
 
+> **ARCHIVED EVIDENCE SNAPSHOT — NOT CURRENT RELEASE AUTHORITY.** This document
+> preserves dated implementation history. Current source, `AGENTS.md`, and the
+> schema-v2 ledger at `mesh/release/readiness.json` control every present-tense
+> release decision. Test counts, SHAs, workflow results, and “complete” claims
+> below must not be reused without fresh same-source verification.
+
 **Prepared:** 2026-07-29
 **Audience:** An autonomous coding agent (and human reviewers) picking up engineering work on Mesh
 **Purpose:** Translate a full-codebase production-readiness review into a prioritized, actionable backlog that moves Mesh from "first successful manual E2E run" to a real production beta.
 
 The machine-readable release gate ledger is [mesh/release/readiness.json](mesh/release/readiness.json), validated by `npm run check:readiness-ledger`. It is release evidence only; runtime security and capability authority remain in the typed backend boundary.
 
-Evidence binding rule: `releaseSha` names the exact source snapshot on which the
-evidence was collected. Because a tracked ledger cannot contain the commit SHA
-of the commit that contains that same ledger, the release workflow permits a
-final metadata-only commit that changes only `mesh/release/readiness.json` after
-the tested source snapshot. The validator rejects any source-code delta in that
-case. Never update the ledger SHA and application code together when preparing a
-release evidence snapshot.
+Evidence binding rule: schema v2 records both `sourceCommit` and
+`sourceTreeHash`; each passing item repeats them as `testedCommit` and
+`testedTreeHash`. A passing ledger therefore identifies one committed source
+tree, while a later metadata-only ledger commit may record that already-tested
+tree. Never treat a dirty worktree or a different tree at the same base commit
+as equivalent evidence.
 
 ---
 

SECURITY.md

@@ -9,18 +9,22 @@ signed public update route is verified.
 
 ## Report a vulnerability
 
-Use GitHub's private vulnerability reporting for this repository. Do not open a
-public issue, discussion, or pull request containing exploit details, account
-tokens, invitation secrets, recovery material, personal data, signing material,
-or service credentials.
+**Confidential route status: unavailable.** GitHub private vulnerability
+reporting is currently disabled for this repository, and no reviewed security
+mailbox has been published. Do not open a public issue, discussion, pull request,
+or advisory draft containing exploit details, account tokens, invitation
+secrets, recovery material, personal data, signing material, or service
+credentials. The production beta is blocked until the owner enables and drills a
+confidential reporting route; this file and the public security page must be
+updated in the same reviewed change when that happens.
 
 Include the affected source SHA, operating system, impact, minimal reproduction,
 and whether the issue may already have exposed data. Do not test against accounts,
 communities, or infrastructure you do not own or have explicit permission to use.
 
-The owner will acknowledge a report, coordinate remediation and disclosure, and
-credit reporters when requested and safe. No response-time SLA is promised before
-the production beta exists.
+Once a confidential route is enabled, the owner will acknowledge reports,
+coordinate remediation and disclosure, and credit reporters when requested and
+safe. No response-time SLA is promised before the production beta exists.
 
 ## Release security boundary
 

mesh/docs/architecture/on-device-ai-boundary.rst

@@ -8,8 +8,8 @@ Required behavior
 -----------------
 
 * Inference stays on the user's device. Production code cannot add a third-party
-  inference endpoint or AI-provider SDK without an explicit security-reviewed
-  allowlist change in ``scripts/check-ai-boundary-lib.mjs``.
+  inference endpoint or AI-provider SDK. The reviewed manifest at
+  ``security/ai-boundary.json`` keeps both network allowlists empty.
 * AI code cannot send messages, create invitations, remove or ban people, change
   roles, or perform moderation. Suggestions must remain drafts that a person
   reviews and submits through the normal product controls.
@@ -27,7 +27,13 @@ Review boundary
 ---------------
 
 The CI check scans production source and dependency manifests, not prose. It
-rejects known provider dependencies and endpoints globally, then applies
+validates the checked policy manifest, rejects known provider dependencies and
+endpoints globally, then applies
 network, feature-gate, download, resource-disclosure, and user-authority checks
 only to narrowly identified AI modules. Positive and negative fixtures keep the
 rule behavior reviewable.
+
+The same manifest is embedded and validated by ``src-tauri/src/ai_boundary.rs``.
+Native authority permits only a draft suggestion and denies network, sending,
+invitation, membership, role, and moderation actions. Renderer markers and the
+source scanner are defense in depth; they never grant authority.