Systems languages orchestrator

Systems Languages Orchestrator

The single entry skill for Go, C++, and Perl work. It locates the task on the language × concern map and delegates to one of 7 specialist spokes. The cross-cutting model every spoke shares — pick the language, then move write → prove → (for Perl) harden, with each language's idiomatic style and toolchain — lives in systems-languages-core; read it before choosing conventions or wiring a test/coverage command.

Cluster map (spoke → role)

Go

• golang-patterns — idiomatic Go: accept-interfaces/return-structs, useful zero values, error wrapping, concurrency (worker pools, context, errgroup), functional options, package layout, performance.
• golang-testing — table-driven tests, subtests, t.Helper/t.Cleanup, golden files, benchmarks, fuzzing (1.18+), coverage, HTTP handler tests — TDD-first.

C++

• cpp-coding-standards — the C++ Core Guidelines distilled: RAII, immutability-by-default, Rule of Zero/Five, smart pointers, enum class, concepts, no C-style casts, naming, performance.
• cpp-testing — GoogleTest/GoogleMock with CMake/CTest: fixtures, mocks vs fakes, gtest_discover_tests, coverage (gcov/llvm-cov), sanitizers (ASan/UBSan/TSan), flaky-test guardrails.

Perl

• perl-patterns — modern Perl 5.36+: use v5.36, signatures, postfix deref, isa, Moo/Corinna OO, named-capture regex, Path::Tiny, Exporter, perltidy/perlcritic/carton tooling.
• perl-testing — Test2::V0 (+ Test::More), prove runner, hash/array/bag builders, subtests, exception tests (dies/lives), Test::MockModule, Devel::Cover — TDD-first.
• perl-security — taint mode (-T), allowlist input validation, three-arg open, list-form process exec, DBI placeholders, web hardening (XSS/SQLi/CSRF), ReDoS, perlcritic security policies.

Routing rules by intent

1. "Which language?" unresolved → read systems-languages-core §1 first, then route by the language axis.
2. Go app/service → golang-patterns (+ golang-testing for tests, benchmarks, fuzzing).
3. C++ code → cpp-coding-standards (+ cpp-testing for GoogleTest/CTest, coverage, sanitizers).
4. Perl code → perl-patterns (+ perl-testing for tests; always add perl-security when the code touches user input, the shell, the filesystem, or SQL — that is Perl's defining hazard).
5. "Write the tests" / TDD → the matching *-testing spoke for the chosen language.
6. "Is it safe?" / handling untrusted input in Perl → perl-security (Go/C++ have no dedicated security spoke here; apply input-validation + RAII/error-handling rules from their patterns/standards spokes).
7. "Review / refactor" → the language's patterns or coding-standards spoke; pair with its *-testing spoke to lock behavior first.

Standard flow

1. Resolve the language (Go · C++ · Perl) — pull the decision model from systems-languages-core §1 if unstated.
2. Apply the language's conventions spoke (golang-patterns · cpp-coding-standards · perl-patterns).
3. Drive it test-first with the matching *-testing spoke (golang-testing · cpp-testing · perl-testing).
4. For Perl that touches input/shell/FS/SQL, harden with perl-security before shipping.
5. Return: chosen language, the spokes engaged, the toolchain/test command, and the next action.

Guardrails

See systems-languages-core. In short: stay in one language's lane — don't carry Go idioms into C++ or Perl conventions into Go; each has its own style, error model, and toolchain. Tests come first (the *-testing spoke), and for Perl, security is not optional when untrusted data is involved — taint mode, allowlist validation, parameterized DBI, and list-form process exec are table stakes (perl-security). Keep Go simple and error-explicit, keep C++ RAII + const-by-default, keep Perl on use v5.36 with safe defaults. Never widen an input surface (new shell call, raw SQL, two-arg open) without saying so explicitly.

Loading spokes on demand

To keep CLI startup context lean, this cluster's spokes are not separately registered as skills — only this orchestrator and its *-core are enumerated. When you route to a spoke named above, load it on demand by reading its file:

~/.agents/skill-clusters/skills/<spoke-name>/SKILL.md (or skills/<spoke-name>/SKILL.md inside the skill-clusters repo).