docs: add comprehensive architecture review & grades for libs subsystem

[sroussey]

Summary

Add a detailed senior-architect review document (prd/analysis/grades/README.md) evaluating the @workglow-dev/libs library (v0.3.7) across all 34 subsystems (13 core packages + 21 provider packages). The review provides letter grades, numerical scores, and actionable recommendations for the entire system.

Key changes

• Subsystem grading matrix: Grades and scores for all core packages (job-queue A/93, util A−/90, workglow A−/90, etc.) and provider packages (AI-inference, infrastructure/runtime, storage backends)
• System-wide assessment: Overall libs grade B+/87, with system-wide grade A−/88 across all three repos (libs, builder, sec)
• Dependency graph documentation: Explicit mapping of how subsystems compose (foundation → engine → AI → providers)
• Cross-cutting friction analysis: Four major system-wide issues identified:
  • Leaky DI isolation (global singletons defeating createOrchestrationContext)
  • any at SDK/dataflow boundaries erasing type safety
  • Capability/transaction guarantees leaked via prose instead of types
  • Security-critical code lacking in-package tests (SafeFetch, JavaScriptTask, MCP auth, etc.)
• Top recommendations: Five prioritized actions to move the system to solid A grade, including making DI isolation real, introducing capability interfaces, driving down any with typed port-value model, hardening untrusted execution, and unifying SQL identifier quoting

Notable details

• Mean subsystem score ≈ 85.5; core engine is genuinely A−/A grade (job-queue, util, storage, ai, knowledge-base all 89–93)
• Provider layer is unusually consistent for 21 packages; score pulled down by mlx stub (D/55), four conformance-skipping providers, and javascript sandbox's missing timeout
• Recurring themes identified: leaky DI isolation, any at boundaries, missing in-package tests on security-critical code, and several stubs shipped alongside working code
• Document serves as both assessment and roadmap for architectural improvements

https://claude.ai/code/session_01AmWzqVKXTNzNtdrSx9fTok

[github-actions]

Coverage Report

Status	Category	Percentage	Covered / Total
🔵	Lines	62.11%	24562 / 39540
🔵	Statements	61.97%	25415 / 41006
🔵	Functions	63.03%	4656 / 7386
🔵	Branches	50.75%	12027 / 23698

File Coverage

No changed files found.

Generated in workflow #2436 for commit cf0af4f by the Vitest Coverage Report Action