Freeze Bible v19.0 as authoritative framework for all future MCP serve…
Decision
Freeze Bible v19.0 as authoritative framework for all future MCP server and paired skill development, on branch v19.0 at commit 2057aa5, strictly additive to v18.1 (no existing rules removed or modified), with Snowflake pair selected as the first pilot.
Rationale
(1) Research-backed: three parallel audit streams established ground truth — Bible v18.1 state (28 docs, Python-only, 7 Laws, targeting MCP spec 2025-11-25), MCP ecosystem currency (spec 2025-11-25 current, Python SDK 1.27.0 latest, v2 alpha banned, 5 high-value bumps identified including pytest CVE-2025-71176), VPS fleet reality (60 MCP servers, 22 Redis sidecars with 3 zombies + 9 silent-degraders, 2 tight skill-MCP pairs out of ~40 possible). (2) The “MCP is dead” article critiques were evaluated against AJ’s actual setup: token bloat mitigated by on-demand Connector tier, complexity mitigated by standardized scaffold, security mitigated by Traefik+Platform-Managed auth — MCP remains the right bet; CLI+API prescription is regression. (3) Core gap identified via Snowflake deep-dive: 856 tools including cost-governance (estimate_query_cost, create_resource_monitor, calculate_credit_burn_rate) exposed but QUERY_TAG appears zero times — the tools exist, enforcement does not, and no contract binds skill expertise to server enforcement. (4) v19.0 addresses this with three new Inviolable Laws (Pairing Contract, Operational Guardrails, Drift Detection), a 12-section Pairing Contract template, a 3+1-layer Mandatory-Invocation Mechanism (frontmatter + hookify autoloader + drift cron + aspirational server header), an 11-guardrail Operational Catalog, Apr-2026-current stack pins with Chainguard + supply-chain trio, and a 9-phase zero-disruption upgrade playbook. (5) Snowflake pilot selected on impact = tool-count × usage × cost-exposure; Contract fully resolved as reference at pairing-contracts/snowflake.md. (6) User approved direction explicitly; framework committed to v19.0 branch (not pushed — awaiting explicit push authorization per standing rule). (7) Known gap acknowledged: Phase 0 verification tooling (~10 scripts + 1 Hookify rule) not yet authored — required before Snowflake Phase 1 can begin; tracked transparently in commit trailer Not-tested field.
Alternatives Rejected
Outcome
Pending
Related
- docker
- oracle
- snowflake
- 2026-04-04-oracle-001-self-architecture-analysis
- graphiti
- execute-v190-thoroughness-pass-as-commit-6d3760e-expand-bibl
- execute-deep-quality-audit-pass-commit-7df0512-fixing-one-p0
- v191-gap-closure-pass-complete-commit-0af70cf-three-new-docu
- v1911-completion-operational-skill-sync-1-added-13upstreamsy
- fivetran-phase-1-findings-l4-single-stage-dockerfile-l5-no-n
- r4-hookify-autoloader-consumer-hook-missing-is-a-canonizatio
- build-g2-cost-pre-flight-as-runwal-local-guardrail-inside-fi
- blue-capability-stack-hybrid-mcp-primary-ai-surface-cli-v066
- proceed-to-phase-3-data-layer-with-three-explicit-bias-corre
- researchclaw-v040-ship-14-tool-streamable-http-merged-server
- salesforce-mcp-skill-upgrade-scoped-to-sales-cloud-service-c
- park-19-salesforce-agentforce-tools-via-skillmemory-only-lig
- author-snowflake-mcp-skill-upgrade-plan-to-v2200-bible-v1917
- snowflake-mcp-v2202-signed-at-880-honest-tools-after-r38-r40
- bible-v181-v1917-framework-upgrade-for-tableau-mcp-pair-ship
- of-the-3-deferred-risks-from-the-dual-stack-browser-install
- architectural-rule-for-helix-going-forward-absorb-data-rules
- adopt-registry-dispatch-as-a-third-architectural-pattern-r50