RetroAgile Workspace

Multi-Project Intelligence

Compare projects side-by-side, trace cross-repo blast radius, inspect code & error heatmaps, and catch systemic rework patterns.

Repositories Grouped by Project (4 Active Services)

Select a project to filter dependency graphs, diagnostics, heatmaps, and 72-hour code survival.

Live Multi-Project Dependency Topology

Shared Contracts & Cross-Repo Blast Radius

Animated streams highlight runtime dependencies, contract coupling, and downstream impact when shared types change.

Clean Flow Churn / Drift Edge Shared Domain Contracts
Multi-Project Trends & Footprint
Tracks 1st-pass shipped PRs alongside 72-hour code survival and rework spend across sprints.

Engineering Intelligence & Systemic Diagnostics

Continuously surfaces chronic file hotspots, 72h shadow rework, agent-model mismatches, test mocking shortcuts, review bottlenecks, and cross-repo contract risks.

Architecture & Hotspotsfintech-ledger
86% PR touch rate · $11.40 churn

src/services/billing.ts is modified in 86% of PRs — split the god-module

Across 6 unrelated features, agents had to edit the same 1,100-LoC file coupling currency rounding, tax tables, PDF export, and SQL queries.

Current Hotspot Impact19 edits · 6 reverts

Agents lose context across mixed concerns and trigger circuit breakers inside billing.ts.

Architectural OpportunityExtract MoneyMinorUnit

Isolate pure minor-unit rounding and tax matrix fixtures into @acme/money-domain.

Recommendation: Break billing.ts apart at the module boundary rather than patching agent prompts every sprint.
Code Survival & ReworkCross-Repo
66% vs 97% 72h survival

34% of merged fintech agent code is rewritten by humans within 72 hours

Low 72-hour survival exposes 'shadow rework': PRs pass green on merge day, then senior engineers quietly rewrite subtle concurrency and rounding bugs in production.

fintech-ledger (Shadow Rework)66% 72h survival

184 of 540 merged AI lines touched again in hotfix commits within 3 days of merge.

checkout-api (High Survival)97% 72h survival

Boundary assertions and failure-branch tests catch edge cases before PR approval.

Recommendation: Track post-merge survival by module and flag PRs that ship green tests without negative/edge-case coverage.
Agent & Model FitCross-Repo
3.4× cost/PR delta

Agent model routing mismatch: heavy reasoning spent on UI, fast model looping on state

Your storefront repo spends frontier-tier tokens on simple component wiring while complex ledger migrations run on lightweight autocomplete.

web-storefront · Over-Provisioned21k tok / simple PR

Reasoning model used on bounded UI copy & single-endpoint additions where fast tier ships in 90s.

fintech-ledger · Under-Provisioned2 loops on state logic

Lightweight model loses track of multi-currency invariants across 4+ linked domain files.

Recommendation: Route multi-file state machines to deep-reasoning models and cap single-file UI tasks at the fast tier.
Cross-Repo ContractsCross-Repo
2 downstream repos impacted

Shared DTO edit in checkout-api broke downstream ledger builds 4 hours later

An agent updated InvoiceLineDTO in @acme/shared-contracts for checkout-api without running contract tests against fintech-ledger consumers.

checkout-api (Producer PR #89)Passed repo CI

Local unit suite was green because mocks matched the new DTO shape inside checkout-api.

fintech-ledger (Consumer Break)Failed deploy gate

Downstream ledger worker still expected legacy float field taxAmount until manually patched.

Recommendation: Require cross-repo consumer contract checks whenever an agent modifies exported types under @acme/shared-contracts.
Test & CI Integrityevent-pipeline
False-green CI detected

Agent mocked the schema validator inside batch-writer tests to make CI pass

Instead of handling malformed webhook batches in batch-writer.ts, an AI session stubbed validateBatch() inside Vitest—shipping a false-green PR.

Detected Pattern in PR Diffvi.mock('../validator')

Added mock bypassing 400 error assertion after initial Vitest failure on malformed spans.

Root Fix via Adapter Registry58% PR churn -> 0

Extract agent payload adapters into a declarative table tested with real unmocked fixtures.

Recommendation: Block tests that mock core domain validators and refactor batch-writer.ts onto a declarative adapter registry.
Cross-Repo ContractsCross-Repo
ACR 0.00 vs ACR 4.75

Money arithmetic diverges across services: BIGINT minor units vs floating-point

checkout-api enforces integer minor units (BIGINT) with RFC-7807 errors and achieves 0.0 ACR, while fintech-ledger still uses decimals and unstandardized strings.

fintech-ledger Pattern$8.42 loop burn

Ad-hoc float rounding triggers cent discrepancies across EUR/RON/GBP invoices.

checkout-api Standard100% 1st-pass CI

Strict minor-unit types & explicit 409/422 error contracts keep AI outputs deterministic.

Recommendation: Promote checkout-api's minor-unit rule into org-wide .cursorrules / CLAUDE.md guardrails.
Review & FlowCross-Repo
74m wait vs 11m fast-track

PR review wait jumps 6.4× when agent changes aren't split into atomic diffs

Agents write code in 3–6 minutes, yet delivery stalls for over an hour whenever diffs mix schema migrations, domain logic, and test fixes.

Large Mixed PRs (fintech #141)74m human wait

400+ LoC mixed-author PR required 3 manual rescue commits before approval.

Atomic Fast-Track (storefront)11m human wait

Scoped <=150 LoC single-pass PRs auto-receive verification receipts and merge 6× faster.

Recommendation: Split multi-concern tickets in Spec Gate before dispatch and fast-track low-ACR single-pass PRs.
Architecture & Hotspotsfintech-ledger
71% endpoint touch rate

Per-route tenant JWT parsing in rls-claims.ts touched on 71% of API endpoints

Every new API route duplicates JWT tenant claim extraction and manual isolation checks instead of wrapping handlers in @acme/auth-rls middleware.

Manual Copy-Paste Pattern11 edits · 3 reverts

Agents forget cross-tenant 403 assertions whenever tenant checks live inside each route.

Centralized @acme/auth-rls1 shared wrapper

Move tenant extraction into withTenantContext() so new endpoints inherit isolation automatically.

Recommendation: Generate a high-clarity refactoring spec to replace per-route auth boilerplate with a shared middleware wrapper.

Code Churn, 72h Rewrite & Failure Mode Heatmaps

Identify files rewritten repeatedly by agents or humans, and see where quality failures cluster by repo and tool.

Code Modification & Churn Heatmap

Edits & reverts per file across Sprint 25 checkpoints

Error, Test Mock & Failure Mode Heatmap

Incident clusters across repositories, coding agents, and failure classes