engineering-verification-loop

Top-level engineering verification orchestration skill and distributable skill-pack entry point. Use when Codex must plan, execute, or review a Go/Python-heavy engineering change that needs an auditable TDD inner loop plus risk-based gates for formal correctness, model-to-implementation conformance, algorithms, performance, API/schema compatibility, CI failures, profiling, distributed behavior, or experiment claims.

Algorithm and data-structure selection specialist for engineering-verification-loop. Use directly or when routed by engineering-verification-loop to choose, replace, or evaluate algorithms or data structures under latency, throughput, memory, asymptotic-complexity, or benchmark-regression constraints, especially when a Dafny-backed correctness model exists but the target Go/Python implementation must be efficient. Not for formal proof, distributed protocol semantics, network partitions, clocks, consensus, broad infrastructure tuning, or generic refactoring.

API and schema compatibility specialist for engineering-verification-loop. Use directly or when routed by engineering-verification-loop to check OpenAPI, JSON Schema, Protobuf/gRPC, GraphQL schemas, HTTP endpoints, event payloads, SDK interfaces, or serialized wire formats for breaking changes against a baseline; not for endpoint implementation correctness, auth proof, distributed protocol semantics, performance, or general API documentation.

CI failure and regression forensics specialist for engineering-verification-loop. Use directly or when routed by engineering-verification-loop to parse structured CI failure footers, reproduce failing CI locally, classify build/test/lint/typecheck/performance failures, extract stable failure signatures, compare environments, isolate first-bad commits, or produce a minimal fix plan from logs/artifacts. Not for designing new tests, formal verification, API compatibility review, or performance optimization after the failing regression is isolated.

Dafny formal model verification specialist for engineering-verification-loop. Use directly or when routed by engineering-verification-loop to create, modify, debug, or review Dafny `.dfy` models, prove properties before implementation, repair Dafny verifier failures, or work on authorization/access control, tenant isolation, state machines, parsers, validators, algorithmic correctness, invariants, security-sensitive branching, or target-language code that should be backed by machine-checked Dafny specifications.

Model-to-implementation conformance specialist for engineering-verification-loop. Use directly or when routed by engineering-verification-loop to map Dafny or TLA model properties to Go/Python implementation evidence, audit model-to-code conformance, compare model-exported fixtures or oracle traces against implementation output, or distinguish formal model proof from handwritten implementation alignment.

Property-based, differential, metamorphic, and stateful testing specialist for engineering-verification-loop. Use directly or when routed by engineering-verification-loop to compare optimized Go/Python code against a simple oracle, old implementation, Dafny-modeled behavior, protocol trace, parser/validator spec, or API fixture across generated inputs. Not for ordinary Red-Green-Refactor orchestration, formal proof, benchmark claims, mutation-score policy, API schema compatibility, or CI log triage.

Test-driven development specialist for engineering-verification-loop. Use directly or when routed for Go/Python features, bug fixes, behavior-preserving refactors, legacy characterization, or behavior-affecting API, CI, performance, formal-correctness, and distributed changes that need auditable Red-Green-Refactor evidence, requirement-to-test traceability, regression scope, coverage, mutation, and flake controls. Not for formal proof, API compatibility analysis, profiling, distributed model checking, or CI log forensics.

Profiler-guided performance optimization specialist for engineering-verification-loop. Use directly or when routed by engineering-verification-loop to optimize latency, throughput, CPU, memory, allocation, I/O, startup time, or hot-path runtime behavior only after identifying measured bottlenecks with profiler evidence and to verify before/after performance after editing.

Reproducible experiment and benchmark analysis specialist for engineering-verification-loop. Use directly or when routed by engineering-verification-loop to audit experiment results, benchmark reports, ablations, model metrics, algorithm comparisons, rollout measurements, or research claims that need reproducible commands, fixed inputs, captured environments, seeds, repeated runs, statistical uncertainty, workload-validity evidence, or falsifiable conclusions.

TLA+/PlusCal distributed model-checking specialist for engineering-verification-loop. Use directly or when routed by engineering-verification-loop to model or check safety/liveness under asynchronous message ordering, retries, duplication, loss, network partitions, quorum behavior, leader election, replication, clock skew, crash-stop, crash-recovery, or consensus-style nondeterminism. Not for local Dafny proof repair, ordinary single-process state machines, algorithm performance optimization, production chaos testing, or generic distributed-systems advice.