Netcracker

Build the APIHub backend, start it locally with local_config, author or extend a Postman/Newman E2E collection under tests/api covering the change, run it, and analyze and fix failures. Use after implementing a non-trivial backend change, before finishing the task, to verify it end-to-end against a real running instance.

Implements and modifies the APIHub Go backend (qubership-apihub-service): controllers, services, repositories, entities, migrations, Service.go wiring, ErrorCodes, and OpenAPI specs. Use when adding or changing backend features, REST endpoints, SQL migrations, repository queries, or Go code in qubership-apihub-backend.

Reviews APIHub backend code changes against project standards (AGENTS.md, rules, development guide). Use when the user asks for self-review, code review of a diff, or post-implementation check before commit or PR in qubership-apihub-backend. Invoke explicitly after another agent or model wrote the code.

Implements and modifies the APIHub Go backend (qubership-apihub-service): controllers, services, repositories, entities, migrations, Service.go wiring, ErrorCodes, and OpenAPI specs. Use when adding or changing backend features, REST endpoints, SQL migrations, repository queries, or Go code in qubership-apihub-backend.

Implements and modifies Go backend services in the APIHub ecosystem (controllers, services, repositories, entities, migrations, wiring, and OpenAPI). Use when adding or changing backend features, REST endpoints, SQL migrations, repository queries, or Go code in APIHub Go repositories.

Reviews Go backend code changes against project standards (AGENTS.md, rules, development guide). Use when the user asks for self-review, code review of a diff, or post-implementation check before commit or PR. Invoke explicitly after another agent or model wrote the code.

Plan implementation from a GitHub ticket by clarifying missing requirements, asking explicit questions for ambiguities, analyzing the codebase, and producing a concrete execution plan. After manual approval, publish the approved plan as a comment on the GitHub issue. Use when the user asks to plan work from a GitHub issue, ticket, or task description.

Decide whether a backend change requires follow-up in qubership-apihub-postman-collections (Newman E2E) and remind the developer with concrete actions. Use after REST or OpenAPI contract changes in the backend — not when authoring collections inside the Postman repo itself.

How to author APM packages — instructions, skills, prompts, agents, hooks. Use when editing APM primitives (`apm.yml`, `*.instructions.md`, `SKILL.md`, `*.prompt.md`, `*.agent.md`, hooks) or laying out a new `agent-packages/<name>/`.

Run a code review through the Codex CLI and apply the fixes — Codex reports findings, you classify each one (real issue vs false positive), fix the real ones, and re-review until the review is clean. Use only when the user asks for a Codex review (cross-review or second opinion) of a branch diff, the uncommitted changes, or a specific commit.

American-English-first style guidance for developer-facing text of any length — README and docs, code comments and docstrings, Javadoc / KDoc / TSDoc / JSDoc, commit messages, PR descriptions and review replies, design docs, changelog entries, UI strings (buttons, labels, placeholders, tooltips), error and log messages, and English-side localization files (.po, .properties, JSON i18n). Use whenever the task involves English developer text — writing, editing, rewriting, translating, localizing, reviewing, proofreading, verifying, auditing, double-checking, or cross-checking. Length does not matter: a one-line `msgstr`, a `// FIXME` comment, or a three-word button label is in scope. Also covers choice of English identifier names (methods, variables, classes, files, tests) read by humans. Covers em-dash usage, hyphen stacking, sentence structure, that/which, gerunds after rather than, hedging, inclusive language, and British/American dialect consistency. Do not use for casual chat replies to the user.

Apply Markdown drafting rules an autofixer cannot fix — wrap body lines at 120 chars (MD013), one H1 per document (MD025), named links instead of bare URLs (MD034), language tag on every fenced code block (MD040). Use when editing `*.md` in a repository that pins markdownlint `MD013.line_length` to 120.

Navigation-only skill for individual actions in netcracker/qubership-workflow-hub. Use when a workflow needs to consume a specific Qubership action (Docker build/push, version/tag rendering, Maven/npm/Python publishing, package cleanup, Helm charts, security scans, etc.) and you need to find the right action and read its authoritative README. All rules (pinning, permissions, anti-hallucination, naming) live in qubership-workflow-conventions — this skill does not restate them.

Navigation-only skill for the Netcracker workflow-templates catalog at Netcracker/.github/workflow-templates. Use when the user asks for a CI/CD workflow that may already exist as a curated template (Docker build/release, Maven/npm/Python release, Helm chart release, security scan, SBOM, PR hygiene, license/lint, GHCR cleanup). Points to the canonical template files and explains how to fork. All rules (pinning, permissions, anti-hallucination, naming) live in qubership-workflow-conventions — this skill does not restate them.

Single source of truth for Qubership GitHub Actions workflows. Use when designing, writing, reviewing, or debugging .github/workflows/*.yml that consume actions or templates from netcracker/qubership-workflow-hub or Netcracker/.github.

Navigation-only skill for individual actions in netcracker/qubership-workflow-hub. Use when a workflow needs to consume a specific Qubership action (Docker build/push, version/tag rendering, Maven/npm/Python publishing, package cleanup, Helm charts, security scans, etc.) and you need to find the right action and read its authoritative README. All rules (pinning, permissions, anti-hallucination, naming) live in qubership-workflow-conventions — this skill does not restate them.

Markdownlint ruleset from super-linter. Apply when writing or editing any .md file.

Single source of truth for Qubership GitHub Actions workflows. Use when designing, writing, reviewing, or debugging .github/workflows/*.yml that consume actions or templates from netcracker/qubership-workflow-hub or Netcracker/.github.

Audit GitHub Actions workflow and action.yml files for security issues using zizmor rules, then fix violations

Update or create documentation for a specific action or reusable workflow based on git diff and action.yml

Audit files for EditorConfig violations (final newline, trailing whitespace, indent, line endings) and fix them in-place

Generate a PR title and description following project conventions, then create or update the PR via gh CLI

Prepare a pull request that adds Qubership baseline agent support to a repository: install qubership-essentials, run apm install for repository-declared harnesses, commit generated agent assets, and configure generated-file attributes and Super-Linter exclusions.

Author an Architecture Decision Record. Use whenever the user wants to capture, write, or draft an architecture or system-design decision — "write an ADR", "document this decision", "add a decision record", "we decided to use X, record it" — even when they don't say "ADR" but describe a choice worth tracking (a pattern, a non-functional requirement, a dependency, an interface, or a tool selection). Also use to supersede or amend an existing ADR.

British-English-first style guidance for developer-facing text of any length — README and docs, code comments and docstrings, Javadoc / KDoc / TSDoc / JSDoc, commit messages, PR descriptions and review replies, design docs, changelog entries, UI strings (buttons, labels, placeholders, tooltips), error and log messages, and English-side localisation files (.po, .properties, JSON i18n). Use whenever the task involves English developer text — writing, editing, rewriting, translating, localising, reviewing, proofreading, verifying, auditing, double-checking, or cross-checking. Length does not matter: a one-line `msgstr`, a `// FIXME` comment, or a three-word button label is in scope. Also covers choice of English identifier names (methods, variables, classes, files, tests) read by humans. Covers em-dash usage, hyphen stacking, sentence structure, that/which, gerunds after rather than, hedging, inclusive language, and British/American dialect consistency. Do not use for casual chat replies to the user.

Triage failing checks on dependency-update PRs (Renovate or Dependabot) and drive them back to green: classify each failure, fix the clear ones, ask about the rest. Use only when the user asks to triage, fix, or sweep failing checks on dependency-update PRs — pass a PR number for one PR, or none to process every open dependency-update PR with a failing check.

Set up Renovate so dependency-update PRs merge themselves once CI passes. Use when enabling automerge in renovate.json, deciding which GitHub status checks to mark as required so a green PR means the tests actually ran, or adding a CI-gate job that folds several test jobs into one required check.

French technical-prose conventions for software documentation, README files, UI strings, error messages, logs, code comments, PR descriptions, changelog entries, and French localization. Use when writing, editing, translating, or reviewing French software-facing prose. Defaults to fr-FR, supports fr-CA mode, and focuses on clear natural French, technical precision, terminology consistency, and avoiding over-translated or LLM-like phrasing.

Russian style guidance for developer-facing text of any length — README and docs, code comments, docstrings, commit messages, PR descriptions, changelog entries, UI strings (buttons, labels, placeholders, tooltips), error and log messages, and localisation files (.po, .properties, JSON i18n). Use whenever the task involves Russian developer text — writing, editing, rewriting, translating, localising, reviewing, proofreading, verifying, auditing, double-checking, sanity-checking, cross-checking. Trigger verbs include the Russian "напиши", "переведи", "локализуй", "проверь", "перепроверь", "сверь", "отревью". Length does not matter: a one-line `msgstr` is in scope. Focuses on clear natural Russian, technical precision, terminology consistency, and avoiding bureaucratic or LLM-like phrasing.

Mapper Element provides an ability to setup a mapping between the message from source element to the target element. Use this skill when performing mapping between two elements, requiring data transformation, the user asks to add mapper/mapper-2.

API integration and utility usage conventions in QIP UI. Use when editing src/api, src/misc, or code that consumes these layers.

Component implementation patterns for QIP UI. Use when editing tables, forms, modals, notifications, labels, icons, and permissions-related UI.

Core development standards for the Qubership Integration Platform UI. Use when implementing, refactoring, or testing code in ui/.

Theme and environment behavior for QIP UI. Use when working on styles, color logic, rendering, or mode-specific behavior.

Convert Qubership Cloud-Core Mesh Helm CRs (FacadeService, Gateway, RouteConfiguration) to Istio Ambient Mesh Gateway API resources (Gateway + HTTPRoute), keeping the chart deployable on both mesh types. Use when asked to migrate, convert, or transform mesh CRs in a Helm chart to Istio / Gateway API.

Orchestrate the full Cloud-Core Mesh to Istio migration end-to-end — convert declarative mesh CRs to Gateway API, migrate route-registration libraries, wire SERVICE_MESH_TYPE, add the Java HTTPRoute generator, generate HTTPRoutes from Go/Java code, validate Istio guards, and maintain MIGRATION_LOG.md. Use when asked to migrate a service from Core Mesh to Istio (Ambient Mesh) or run the migration guide end-to-end.

Generate Gateway API HTTPRoute CRs from Go or Java route-registration code (routeregistration.Route / RouteEntry call sites). Use when asked to generate HTTPRoutes from source code, convert route registrations to HTTPRoute YAML, or extract routes from Go/Java files.

Shared procedure for sorting Gateway API HTTPRoute `rules[]` by path specificity (most specific first). Not triggered directly by users — invoked as a sub-procedure by other skills (`httproute-from-code`, `core-mesh-crs-to-gatewayapi`) before they emit an HTTPRoute. Apply when ordering HTTPRoute rules so longest/most-specific prefix matches appear first.

Authors and updates APIHUB deployment assets in qubership-apihub: Helm values/templates, Docker Compose env wiring, docs/configuration-reference.md parity, local-k8s-quickstart, and optional extensions (linter, agents-backend). Use when changing helm-templates/, docker-compose/, or deployment docs under docs/.

Decide whether a backend change requires follow-up in the qubership-apihub deployment repo (Helm, Compose, configuration reference) and remind the developer with concrete actions. Use when editing backend config, env defaults, or deployment-facing behaviour — not when authoring chart changes inside qubership-apihub itself.

British-English-first style guidance for developer-facing text of any length — README and docs, code comments and docstrings inside source files (.go, .js, .ts, .py, .java, .rs, .kt, .cs, .cpp, .rb, .swift, ...), Javadoc / KDoc / TSDoc / JSDoc, commit messages, PR descriptions and review replies, design docs, changelog entries, UI strings (buttons, labels, placeholders, tooltips), error and log messages, and English-side localisation files (.po, .properties, JSON i18n). Use whenever the task involves English developer text — writing, editing, rewriting, translating, localising, reviewing, proofreading, verifying, auditing, double-checking, sanity-checking, cross-checking, or "checking the wording". Length does not matter: a one-line `msgstr`, a `// FIXME: …` comment, or a three-word button label is in scope. Also covers choice of English identifier names (methods, variables, classes, files, tests) that will be read by humans. Covers em-dash usage, hyphen stacking, sentence structure, that/which, gerunds after

Navigation-only skill for individual actions in netcracker/qubership-workflow-hub. Use when a workflow needs to consume a specific Qubership action (Docker build/push, version/tag rendering, Maven/npm/Python publishing, package cleanup, Helm charts, security scans, etc.) and you need to find the right action and read its authoritative README. All rules (pinning, permissions, anti-hallucination, naming) live in qubership-workflow-conventions — this skill does not restate them.

Single source of truth for Qubership GitHub Actions workflows. Use when designing, writing, reviewing, or debugging .github/workflows/*.yml that consume actions or templates from netcracker/qubership-workflow-hub or Netcracker/.github.

Upgrades Apache NiFi to a target version. Updates scripts, configs, Dockerfile, pom.xml, and checks migration guidance.

Use when adding or changing crawl/clone/transform behaviour, hook lifecycle, path rules, or exported utilities inside the json-crawl library.

Use when adding or changing json-crawl unit tests — crawl order, hook lifecycle, path rules, and clone/transform assertions.

Use when consuming json-crawl from another TypeScript project — crawling or cloning JSON trees, wiring path rules, or reading hook lifecycle semantics.

Use when adding or changing types, factories, parser handlers, or resolution logic inside the ddlapi library — the rules that keep the model faithful to its Atlas-Go origin and the parser's two-pass contract intact.

Use when adding or changing ddlapi tests — choosing the right test file, adding a SQL fixture, asserting on parsed Realms or escape-hatch attrs, or extending the compile-time exhaustiveness check.

Use when consuming the ddlapi library from another TypeScript project — parsing DDL with buildFromDdl, walking the Realm/Schema/Table model, reading PostgreSQL-specific details off the escape hatch, or building schema objects with the factories.

Migrate Qubership DBaaS-backed Go, Spring, and Quarkus microservices from runtime REST provisioning to deployment-time InternalDatabase and DatabaseSecretClaim resources with mounted Secrets. Use only when the user asks to migrate an existing service to this mounted-secret workflow.

Migrate legacy Core Operator DBaaS declarations to dbaas-operator resources. Use only when the user asks to migrate DatabaseDeclaration/DbPolicy resources to InternalDatabase/DatabaseAccessPolicy.

First-hop classification and localization of incoming Qubership logging-stack support tickets. Use when triaging what a ticket author is asking for (a problem on an installation, a consultation, a feature request, or a security/vulnerability remediation) and where it sits (which component, on VM or Kubernetes, at runtime or during deploy, and the specific symptom or topic). Decides from ticket content; uses the author's Ticket Type only as a tie-breaker and ignores Resolution/Root Cause. Research-phase skill; classifies only, takes no system action and answers nothing.

First-hop disposition of a classified Qubership logging-stack support ticket. Use after logging-l1-classification has labeled a ticket, to decide the next action — ask the author for missing data (Additional Info Required), flag a suspected known issue with a draft reply (Suspected Known Issue), or hand a structured packet to L2 (Handoff to L2). Reads ticket content and attachments (logs, screenshots); never touches live systems, mutates the ticket, or closes it.