Central agentic standards, guidelines, skills & bootstrap tooling for AI coding agents (Claude, Codex, Copilot).
This repository also configures its own agents. See what agent-compass tells them →
Skill Claude CodeCodex
Convert a Figma design-token JSON export into a design.md design-system file (google-labs-code/design.md format). Use when the user has Figma tokens/variables JSON and wants a machine-readable design.md with colors, typography, spacing and rounding tokens.
Skill Claude Code
Documentation generator. Analyzes a module's structure and generates README.md and DESIGN.md skeletons. Use when the user mentions generating docs, creating a README, creating a DESIGN doc, doc skeletons, or doc templates. Auto-triggers when a new module is started.
Skill Claude CodeCodex
Read the project specifications and write one RAID register file per open question, risk, assumption, issue or dependency. Use when the user asks to harvest the open questions, collect the assumptions behind an estimate, fill the RAID register, import the spec-kit Clarifications sessions, or find what the specs leave…
Skill Claude CodeCodex
Analyse what changed since a base git reference and produce the developer action plan — technical impacts, blast radius, and the exact project documents that are now stale. Use when the user asks for the impact of a change, what a branch breaks, what a diff affects, which docs to update after a change, or a technical…
Skill Claude CodeCodex
Produce a detailed, production-ready, plug-and-play implementation spec for a scope of work-items that a developer or agent can start or one-shot: exact files, schema, contracts, logic, resilience/security, jobs, env scaffolding, and tests — including mock/fail-closed paths for anything missing credentials. Use when…
Skill Claude CodeCodex
Turn a LikeC4 model into an OpenSpec change proposal (proposal + spec deltas + tasks). Use when the user wants to derive specs or requirements from a C4 model, or mentions likec4/c4 plus openspec.
Skill Claude CodeCodex
Run broad, multi-phase, autonomous coding work safely: intake, plan, checkpoint, optional one-at-a-time subagents, implementation loops, validation, and final sweep. Use when the user asks to keep going until a large goal is finished or gives many phases.
Skill Claude CodeCodex
Use when the work is native Android or native Apple-platform code — Kotlin, Jetpack Compose, Gradle/AGP, Android manifests and intents, R8, Swift, SwiftUI, UIKit, SwiftData, Xcode, an iOS/iPadOS/watchOS/visionOS target, an emulator or a simulator. Routes the task to the pinned vendor skill that holds current platform…
Skill Claude CodeCodex
Turborepo + NestJS + Drizzle + BullMQ monorepo scaffolding — project structure, shared types, workspace config, Docker, CI/CD, and new feature bootstrapping.
Skill Claude CodeCodex
NestJS module architecture patterns — controllers, services, repositories, DTOs, guards, mappers, and module registration for production APIs.
Skill Claude CodeCodex
Use when the task is DevOps, cloud, Kubernetes, containers, CI/CD, observability, security scanning, secrets, hardening, networking, storage, databases, incident response, LLM/GPU serving, or a compliance framework (SOC 2, ISO 27001, GDPR, HIPAA, PCI DSS, FedRAMP). Installs the pinned operational skill for the task…
Skill Claude CodeCodex
Deep PR/MR review governance: compare code to product/spec docs, security, repo rules, tests, docs sync, and UI evidence before writing a summary and inline comments. Use when asked to review a pull request, merge request, branch, or posted review comments.
Skill Claude CodeCodex
Streamline PR creation, PR reviews, and implementing submitted review fixes with GitHub CLI.
Skill Claude CodeCodex
Audit how much of a project is actually built by diffing its specs, functional docs, and requirement registers against the real code into an honest, verified progress matrix (per feature/subfeature and global). Use when asked how far along a project is, what's left, percent complete, a progress or status/completion…
Skill Claude Code
Use when a project has projectmem or durable agent memory. Reads memory before work, checks pre-action warnings, and records durable findings, failed attempts, decisions, fixes, validation, and risks.
Skill Claude CodeCodex
Run an end-to-end QA review pass and write the tester report — every requirement scenario given a verdict, every screen compared against its design reference, every finding split into a defect or a change request. Use when a QA tester, a Product Owner or a project manager asks whether the build matches the…
Skill Claude CodeCodex
React admin dashboard patterns — TanStack Router, MUI Material, MVVM ViewModels, RBAC permissions, Keycloak OIDC, React Query CRUD hooks, feature modules, Vite config.
Skill Claude CodeCodex
Resilience (circuit breaker, retry) and observability (OpenTelemetry tracing, metrics, structured logging) patterns for NestJS production APIs.
Skill Claude CodeCodex
Triage a specification change against the tickets it already produced — summarise the delta in the vocabulary of the ticket, say what to do about it, and flag whether it is a candidate for a change request. Use when a spec moved after tickets were created, when the ticket registry reports drift or a conflict, or when…
Skill Claude CodeCodex
Write the tracker ticket text for one specification section — summary, short why, acceptance criteria taken from the words of the spec, and the link back. Use when the user asks to turn a spec or a requirement into a ticket, to draft a Jira story from a specification, to write acceptance criteria for a requirement, or…
Skill Claude Code
Use for project ideas, feature specs, requirements, unclear tasks, technical plans, task breakdowns, or requests that need spec-driven clarification before implementation.
Skill Claude CodeCodex
Use when refreshing AGENTS.md or provider context markers from current Spec Kit artifacts without overwriting host-owned guidance.
Skill Claude CodeCodex
Use when checking Spec Kit artifacts for inconsistencies across spec, plan, tasks, contracts, docs, and validation.
Skill Claude CodeCodex
Use when creating or updating Spec Kit quality, security, UX, or readiness checklists for a feature spec.
At most 3 mods per repository are shown here, and a mod shipped inside a plugin is left to that plugin's page — the rest are on their repository pages: