“EN-08-AI-Supply-Chain-Adversarial-Eval”

Glossary of domain terms as used in the code, and where the authoritative

"Where would I change X?" map for 5 common changes (add endpoint, add field,

Confusing/dead-looking areas and questions to ask the team — label your

Test-Coverage Gap Analyzer

Identifies untested critical paths (not line coverage) and writes the highest-value missing tests first.

Map behaviors (not lines) to tests: which business rules, branches, and

Rank gaps by risk: money/auth/data-loss/permissions > complex branching >

Identify WEAK tests: no assertions, assertion-free mocks, tests that would pass

Write the top 5 missing tests (boundary + failure paths), table-driven, with

Recommend a realistic coverage policy: critical-package threshold + mandatory

Monolith-to-Module / Service Extraction Planner

Plans extraction of a bounded context from a monolith with data ownership and dual-write/cutover safety.

Bounded-context definition: domain language, owned data (exact tables/columns),

Dependency analysis: inbound/outbound couplings (sync calls, shared DB joins,

Target architecture: in-process module first (seams, interfaces), then optional

Data ownership migration: read model/API replacements for foreign readers,

Strangler steps as independently shippable PRs with feature flags, metrics for

Risks: distributed transactions avoided via saga/eventual consistency;

Code Comment / Documentation Improver

Turns bad comments into maintainable docs: explains WHY, keeps docs near code, and generates API docs from real signatures — no marketing fluff.

Remove/replace comments that restate code; convert every surviving comment into

Docstrings/JSDoc: summary line (imperative), params/returns with units and

A module-level README section: responsibility, when to use/not use, key