[P0][E2E] Prove installed Loop to Runtime MCP execution through Mapper, Dev CLI, watcher and HBP

[wesleysimplicio]

Objective

Provide one release-blocking installed-package E2E proving the complete supported execution chain rather than isolated contracts or checkout-only mocks.

Target chain:

Coordinator -> Simplicio Loop -> Runtime MCP -> Mapper -> Dev CLI/effect executor -> validation -> watcher -> HBP evidence -> Runtime contract verification -> final delivery.

Scope

The test must use released/locally built install artifacts outside all source checkouts. Simplicio Loop remains usable without Runtime, but the Runtime-backed profile must prove that every effect travels through Runtime and no direct mutation bypass occurs.

Implementation steps

1. Build/install pinned Loop, Runtime, Mapper and Dev CLI artifacts into an isolated environment.
2. Record versions, package hashes, commits, OS, CPU, RAM and configuration.
3. Create a small fixture repository with one mechanical task and one semantic task.
4. Start one persistent Runtime MCP session through RuntimeBridge.
5. Negotiate capabilities and contract versions.
6. Run intake, Mapper scan/inspect/handoff and execution route.
7. Execute mutation only through the Runtime effect transaction.
8. Validate with focused tests, independent watcher challenge and acceptance anchor.
9. Publish HBP-linked Loop, Mapper, Dev CLI, Runtime and delivery receipts with one causal run ID.
10. Ask Runtime to consume/verify simplicio.loop-execution/v1.
11. Assert final VERIFIED only when every receipt, hash and watcher condition passes.
12. Repeat with injected failures and upgrade/minimum-supported versions.
13. Store raw metrics and commands; use null with reason for unavailable values.

Mandatory negative lanes

• Runtime missing.
• Runtime explicitly selected but wrong binary.
• Missing Mapper or Dev CLI capability.
• MCP disconnect after effect/before response.
• Corrupt HBP link.
• Stale Mapper artifact.
• Watcher mismatch.
• Duplicate idempotency key.
• Direct file mutation bypass attempt.
• Version/schema mismatch.
• Cancellation and restart.

Acceptance criteria

• Installed artifacts complete the full chain without source-checkout imports.
• Exactly one persistent Runtime process serves the workspace.
• Process trace proves zero cold self-exec on the normal hot path.
• Every mutation has EffectTransaction, gate, validation and HBP receipt.
• Runtime verifies the Loop receipt as VERIFIED.
• Any missing/corrupt/stale link prevents VERIFIED.
• Runtime-optional standalone Loop behavior remains separately tested.
• Linux, macOS and Windows fixtures are covered or unavailable platforms are explicitly reported.
• Raw p50/p95 latency, tokens, CPU/RSS, process count and receipt bytes are published.
• No paid GitHub Actions is required; the suite runs locally/containerized.
• Release cannot claim integrated Runtime support until this lane passes.

Relations

• Runtime Loop consumer: wesleysimplicio/simplicio-runtime#3459
• Runtime conformance: wesleysimplicio/simplicio-runtime#3289
• Runtime warm path: wesleysimplicio/simplicio-runtime#2983

[wesleysimplicio]

Claimed issue #693 on branch codex/issue-693-loop-runtime-mcp-e2e | CLAIMED worker=Codex GPT-5 run=019f8ed8-84a9-7a93-817d-433beab0e795 attempt=1 base=origin/main@1181a71 worktree=C:\Users\Z0059V7A\m\repos\work\simplicio-loop-issue-693. Goal: implement the installed Loop-to-Runtime MCP E2E/verification slice through Mapper, Dev CLI, watcher and HBP; preserve issue scope and keep #693 open unless all acceptance criteria are verified.

[wesleysimplicio]

Progress receipt: PR #711 merged as 147a21d. Local evidence: py_compile, git diff --check, and 3 installed-chain smoke tests passed. Issue remains open for real installed binary/process E2E evidence.

[wesleysimplicio]

MEASURED|Progress receipt for issue #693: PR #721 (3a6a0d87295c09f027557f5da4251363ff88e420) merged into main as 49dce7cabac222ff754eec6390f49333281f7fc5.

MEASURED|Merged diff adds real installed process verification for Mapper, Dev CLI, watcher, HBP and Runtime, including absolute binary/version/SHA receipts, one persistent Runtime PID with two MCP requests, shared correlation IDs, p50/p95/process/receipt-byte metrics, fail-closed effects, Windows cleanup, and focused tests.

MEASURED|Validation: focused process/runtime suite 6/6; production integration 6/6; coverage 85.78% total (86% per touched module, --cov-fail-under=85); Ruff/format/diff-check and impact audit passed. The captured real report had Mapper READY, Dev CLI READY, HBP READY, Runtime READY (pid=23792, requests=2, process_count=1) and watcher BLOCKED because no watcher challenge existed; effects were not attempted.

UNVERIFIED|Issue remains OPEN. AC6 (full wrong/missing/stale/duplicate/direct-mutation negative matrix) and the independent watcher receipt gate still need broader verification.