nxtg-ai
OrganizationForge Plugin: Zero-dependency governance for Claude Code. 21 commands, 22 agents, 29 skills.
Categories
Indexed Skills (27)
architecture
Clean Architecture, Domain-Driven Design, and backend system-design guidance: organizing a codebase into layers with the right dependency direction, choosing structural design patterns (Repository, Strategy, Observer, Command, DI), applying DDD tactical patterns (entities vs value objects, aggregates), designing REST resources and status codes, persistence (unit-of-work, N+1), caching, domain events, layered error handling, and recording decisions as ADRs. Use when designing a new system or feature, deciding which layer code belongs in, refactoring a tangled/God-object codebase, reviewing coupling and boundaries, shaping an API, or writing an Architecture Decision Record.
claude-code-framework
Authoritative reference for how Claude Code itself works AND how to build for it — CLAUDE.md memory, custom slash commands, subagents, skills, hooks, MCP servers, settings precedence, permissions, plan mode, CLI/headless/CI, worktrees, and multi-platform surfaces. Use when authoring or debugging a plugin component (command/agent/skill/hook), wiring an MCP server, choosing frontmatter fields, configuring settings.json or permissions, figuring out why an agent/skill won't auto-trigger, running headless/CI or worktree-parallel sessions, or answering "how does Claude Code X work / how should I structure this for Claude Code."
codex-framework
Reference knowledge for OpenAI Codex CLI compatibility — how Codex discovers AGENTS.md instructions and .agents/skills so a Forge (Claude Code) project can interoperate with Codex without duplicating agent definitions. Use when a user asks to "add Codex support", make the repo "work with Codex CLI" / "work with both Claude Code and Codex", share governance state between the two tools, or when comparing AGENTS.md vs CLAUDE.md discovery. REFERENCE ONLY — never auto-creates AGENTS.md or .agents/ files.
coding-standards
Cross-language coding conventions — naming, type safety, error handling, async, imports, SQL formatting, git commit format, complexity/size limits, security, and a pre-PR review checklist for Python, TypeScript/JavaScript, Rust, and Go. Use when writing or reviewing code, setting a project's style baseline, answering "what's the convention for X", enforcing types/docstrings, formatting a SQL query or commit message, or running the pre-PR checklist. Language deep dives and tool configs live in reference/.
core-nxtg-forge
Ground truth for the NXTG-Forge platform — the real forge CLI subcommands, the /forge:* slash commands, the .forge/state.json + events.jsonl model, both MCP servers (Rust orchestrator + Node governance), the agent roster, and the security hooks. Use when working inside or on NXTG-Forge (the forge-plugin, forge-orchestrator, or forge-ui repos), when a user asks "what does forge <x> do", how state/checkpoints/drift work, which MCP tool to call, which agent to invoke, or when a task references `forge init/plan/run/status`, `/forge:...`, `.forge/`, governance health, or the two forge MCP servers.
domain-knowledge
NXTG-Forge product domain knowledge — the mission, core concepts (project spec, agent orchestration, state, Clean Architecture, drift/gap analysis, MCP), the real `forge` CLI surface, and the 3-repo workspace layout. Use when onboarding to what NXTG-Forge IS and why, when a spec/task/agent-orchestration question needs product framing, or before describing forge CLI/state behavior (avoids repeating retired Python-prototype commands).
optimization
Measure-first performance optimization playbook — profiling discipline, bottleneck triage (CPU/memory/I/O), caching layers, and the algorithm/DB/frontend/network fix patterns. Use when the user reports something is slow ("dashboard takes forever", "API is laggy", "bundle is 2MB", "memory keeps growing"), asks to profile, benchmark, reduce latency/bundle-size, or fix N+1 queries — and to interpret the output of the /forge:optimize command or the `performance` agent.
verify-governance
Adjudicate a flagged governance concern with evidence, then issue a JUSTIFIED or REVERT verdict. Use when a Forge governance hook logs "GOVERNANCE: N violation(s)" or "Code governance", when a test-vs-implementation mismatch is suspected, when a change might be out of directive scope, or when the user asks to verify/justify a code or test change. Reads the real implementation before ruling — it never guesses.
backend-master-agent
Backend implementation playbook — API endpoints, database models/migrations, auth, query optimization, and the framework-specific traps that break servers in production. Use when implementing or reviewing server-side code: FastAPI / Django / Flask / Express / NestJS / Go / Axum handlers, SQLAlchemy / Prisma / Mongoose / Tortoise models, JWT / OAuth / session auth, bcrypt / argon2 password hashing, rate limiting, N+1 query fixes, async/event-loop bugs, or when a handoff spec asks for "the backend" of a feature.
cli-artisan-agent
Knowledge for designing and building command-line interfaces — argument/flag parsing, subcommand structure, interactive prompts, colored and tabular output, progress feedback, exit codes, shell completion, and CLI UX. Use when adding or redesigning a CLI command, choosing a CLI framework (clap, Click/Typer, Commander, Cobra), fixing an argument-parsing or flag-naming bug, adding an interactive wizard or progress bar, or improving CLI error messages and help text.
agent-development
Author and configure Claude Code subagents — valid YAML frontmatter (name, description, model, color, tools, isolation, memory, skills) plus a system-prompt body. Use when creating a new agent .md, fixing an agent that won't load or won't auto-delegate, choosing a model/color, deciding which tools an agent gets, or reviewing agent frontmatter for invalid/ignored fields.
integration-specialist-agent
Resilient third-party integration patterns — API clients, webhooks, MCP server wiring, retries/circuit-breakers, and idempotency. Use when connecting to an external service (Stripe/GitHub/Slack/S3/etc.), building or verifying a webhook receiver, adding retry/backoff or circuit-breaker resilience, registering an MCP server, or reviewing an integration for signature-verification and rate-limit gaps.
lead-architect-agent
Senior architectural decision-making, system design, and clean-architecture layer review for a feature or service. Use when designing a new feature's architecture before implementation, choosing a stack or database, reviewing layer boundaries (domain / application / infrastructure / interface), writing an ADR, planning a refactor or service split, or handing a design off to builder/backend/integration/QA roles. Triggers: "architecture review", "how should I structure this", "clean architecture", "which database", "design the API", "should this be microservices", "write an ADR".
platform-builder-agent
Platform-engineering playbook — Docker/Compose, CI/CD pipelines, Kubernetes, cloud + IaC, monitoring, and the deployment traps that break shipping. Use when the task is containerizing an app, writing a Dockerfile or docker-compose, building a GitHub Actions / GitLab CI pipeline, authoring Kubernetes manifests (Deployment/Service/Ingress/HPA), choosing a deployment target (VPS vs ECS/Cloud Run vs K8s vs serverless), setting up health checks, secrets, autoscaling, backups, or monitoring/observability.
qa-sentinel-agent
Quality-assurance knowledge for test STRATEGY, coverage analysis, bug detection, and quality gates — the reference the qa-sentinel agent reasons from. Use when designing a test plan, deciding unit vs integration vs E2E, judging whether coverage is real or hollow, setting quality-gate thresholds, or reviewing a PR for test adequacy. Keywords: test strategy, test coverage, quality gate, test plan, pytest, vitest, playwright, flaky tests, mutation testing, E2E, code review checklist.
browser-debugging
Debug the running NXTG-Forge web dashboard by driving a headless Chromium via the Playwright MCP server — read console errors, take screenshots, inspect the DOM accessibility tree, and watch network requests. Use when the dashboard renders blank or broken after a UI change, when the user reports browser console errors, after fixing a React bug (infinite loop, "Maximum update depth", stale state) to confirm it's gone, or when verifying a visual/interactive change (Command Center, Infinity Terminal) in the real running app rather than from tests alone.
ceo-loop
CEO Decision Loop — the ORBIT execution protocol (OBSERVE → REASON → BUILD → INSPECT → TURN) for one iteration of NXTG-Forge autonomous product governance. Preloaded into the nxtg-ceo-loop agent and started by the /forge:ceo-loop command; the ceo-loop-stop.sh Stop hook re-invokes it each iteration until the queue empties or limits hit. Use when running or resuming a CEO-LOOP: classify pending decisions by depth, apply the impact×reversibility matrix, write the append-only decision journal + progress file, run retrograde verification, update trust calibration, and decide continue-or-stop.
crucible-audit
Forensic test-quality audit that detects test fraud — hollow assertions, coverage-omit gaming, mock proliferation, dead/gated tests, untested entry points, and README badge inflation that produce false green signals. Use when asked to "run a CRUCIBLE audit", "check/audit test quality", "find silent failures", "find hollow tests", "audit assertions/mocks/coverage config", "why do tests pass but nothing works", or before any release gate (pre-publish/pre-deploy) when a high test count or coverage % needs verifying against reality.
dev-environment-patterns
Debug and configure local dev-server networking for multi-device access — Vite proxy, env-var URL priority, 0.0.0.0 binding, and WSL2 quirks. Use when API/fetch or WebSocket calls fail from a phone/tablet/other PC ("localhost refused", CORS blocked, NetworkError, WS won't connect), when a server is only reachable from localhost, when a leftover VITE_API_URL breaks dev, or when setting up network access on WSL2.
documentation
Documentation standards and code-to-docs sync conventions for Forge projects — JSDoc/TSDoc annotations, README/CHANGELOG structure, auto-generated API reference, doc-tree layout, and staleness detection. Use when writing or reviewing docs, adding JSDoc to exported functions, structuring a docs/ tree, deciding what to document vs auto-generate, or running the Forge docs commands (/forge:docs-status, /forge:docs-audit, /forge:docs-update).
gemini-framework
Reference knowledge for Google Gemini CLI compatibility. Describes how Gemini CLI discovers context (GEMINI.md), configures settings.json, and overrides its system prompt so Forge can interoperate on a dual-platform project. Use when a user asks how Forge coexists with Gemini CLI, how GEMINI.md / @imports / .gemini/settings.json work, or how to add Gemini support alongside Claude Code. DO NOT create files from this skill — it is reference material only.
git-workflow
Trunk-based git workflow for NXTG-Forge — branching strategy, commit conventions, PR practices, rebase/conflict resolution, and how the plugin's own hooks guard git operations. Use when creating a feature branch, writing a commit message or PR, rebasing onto main, resolving merge conflicts, choosing a merge strategy, recovering from a bad commit/reset/detached HEAD, or when a git push is blocked by a security hook.
owasp-security
OWASP vulnerability detection + remediation for security code review, PLUS the Forge plugin's own security guards (why a "SECURITY BLOCK" / exit-2 fired and the correct fix). Use when reviewing or hardening authentication, authorization, input validation, cryptography, API endpoints, session management, deserialization, SSRF, or file uploads; when auditing a Claude Code plugin / MCP server / agent for excessive agency, prompt injection, or supply-chain risk; or when a Bash/Write/Edit was blocked by a Forge guard (rm -rf, chmod 777, curl|sh, force-push, eval(), os.system, shell=True, string-concat SQL, reading .env/*.pem/*.key). Covers OWASP Top 10:2025, API Security Top 10, Agentic AI (ASI01–ASI10), CWE Top 25, ASVS 5.0.
parallel-execution
Claude Code's two force-multipliers — Plan Mode (explore → plan → wait-for-approval → execute) and Agent Teams (parallel subagents via the Task tool) — plus the canonical forge agent roster and delegation-routing rules. Use when decomposing a multi-file feature, running multi-dimensional codebase analysis, orchestrating parallel build/test/validate pipelines, or deciding which subagent to delegate a task to.
runtime-validation
Catch the bugs a green test suite misses by observing REAL runtime behavior — tail application logs during test/server runs, validate API responses against their schema contracts, and assert data-integrity invariants (density ≤ 1.0, %≤100, referential integrity). Use when unit tests pass 100% but the running app still errors, when a Pydantic/Zod/serde validation error or schema drift shows up only at runtime, when reviewing whether a test suite actually exercises the running system, or when someone says "all tests pass but it's broken", "smoke test", "the mocks lied", or "verify it works live".
skill-development
How to author a Claude Code Agent Skill — SKILL.md structure, routing-optimized descriptions, progressive disclosure, valid frontmatter fields, bundling scripts/reference files. Use when creating a new skill, writing or fixing a SKILL.md, deciding what belongs in frontmatter, splitting an oversized skill into reference files, or debugging why a skill never triggers.
testing
How to WRITE real, maintainable tests in any stack — assertions that fail when the code is broken, the test pyramid and AAA structure, behavior-over-implementation, mocking discipline, fixtures, coverage that means something, and the ship gates (test-count-never-decrease, no self-mocking, prove-RED-first). Language-generic with concrete examples in TypeScript/JavaScript, Python, Rust, and Go. Use when asked to "add tests", "write a test", "improve coverage", "fix a flaky test", "how do I test X", when turning a hollow assertion into a real one, deciding what layer a test belongs in, or when a change must not drop test quality. The construction counterpart to the crucible-detective agent's fraud hunt.
Bio shown is the top-scored skill's repo description as a fallback — real GitHub bios land in a future update.