AgentSkills

# AgentSkills Portable Agent Skills, skillsets, MCP presets, and routing templates for AI coding agents. Reusable AI-agent workflows for builders, operators, and executive teams. ## Why this exists Most AI coding agents improve when they are given reusable domain-specific operating instructions. AgentSkills packages those instructions into installable skillsets for Codex, GitHub Copilot, Cursor, Claude Code, Gemini CLI, and related tools. The portable core is vendor-neutral Markdown skills, YAML skillsets, MCP presets, and routing templates. Host-specific setup lives at the edges in setup guides, templates, and adapter scripts. AgentSkills prioritizes clear output contracts, scenario tests, and reviewable safety boundaries over bulk-generated skill lists. ## Quick install Install a discoverable skill with GitHub CLI: ```powershell gh skill install jeremylongworth-source/AgentSkills ``` Pin a specific skill to the current release: ```powershell gh skill install jeremylongworth-source/AgentSkills game-threejs-development --pin v0.1.1 ``` Install a local Codex skillset from a cloned repo: ```powershell .\scripts\install-skillset.ps1 game-dev ``` ## Available skillsets | Skillset | Best for | |---|---| | `game-dev` | Game design, implementation, QA, launch | | `html5-game-publishing` | Phaser, Three.js, itch.io, browser QA | | `executive-command-center` | Executive briefs, decisions, risks, action tracking | | `founder-fundraising-ir` | Investor updates, fundraising narrative, data rooms, diligence | | `ai-transformation-governance` | AI use cases, pilots, governance, vendors, adoption | | `owner-operator-os` | Owner dashboards, cash-flow triage, pricing, vendors, ops | | `operating-cadence` | Leadership meetings, OKRs, decisions, MBRs, follow-up | | `engineering-delivery` | Issue planning, PR review, regression, release risk | | `support-success` | Support triage, safe replies, escalations, KB drafts, account health | | `revenue-growth` | Account research, pipeline, approved outreach, campaigns, funnel | | `sales-marketing` | GTM, positioning, paid media, lifecycle | | `frontend-product` | React, Tailwind, UX, accessibility | | `product-research` | Feedback synthesis, PRDs, experiments, analytics | | `business-analysis` | Guided discovery, BRDs, stakeholder requirements, traceability | | `analytics-finance` | KPI narratives, variance commentary, close checklists | | `research-validation` | Competitive research, experiments, analytics | | `llm-skill-authoring` | Creating and validating new agent skills | | `agentops-evaluation` | Skill benchmarks, scenarios, before/after reports, scoring | | `skill-security-audit` | Skill threat models, prompt-injection review, safe install checks | | `technical-documentation` | READMEs, API docs, release notes, changelogs, migrations | | `backend-api` | API contracts, schemas, auth flows, service boundaries, backend tests | | `devops-cloud-release` | Deployment, CI, containers, config, rollback, cloud cost, readiness | | `quality-testing` | Unit, integration, E2E, bug reproduction, QA, accessibility, performance | | `data-analytics-bi` | Metrics, dashboards, funnels, cohorts, retention, SQL plans, experiments | | `creator-content-engine` | Content pillars, hooks, scripts, calendars, captions, repurposing | | `creator-brand-deals` | Media kits, pitches, sponsorship packages, rate cards, recaps, renewals | | `creator-analytics-reporting` | Sponsor reports, content analytics, affiliate reporting, dashboards, experiments | | `creator-monetization` | Offer ladders, owned-product launches, pricing, affiliates, dashboards | | `creator-business-ops` | Weekly ops, sponsor pipeline, invoices, production handoffs, asset checks | | `creator-community` | Comment and DM triage, moderation briefs, feedback, polls, engagement plans | | `creator-reputation-risk` | Disclosure, rights, AI-labeling, brand-safety, platform-policy, response review | | `creator-ai-production` | AI media prompt briefs, synthetic characters, provenance, disclosure, review gates | | `all` | Every included skill and MCP preset | ## Who this is for - Developers using Codex, GitHub Copilot, Cursor, Claude Code, Gemini CLI, or similar AI coding agents - Builders who want reusable AI-agent workflows instead of rewriting the same instructions for every project - Teams experimenting with MCP-backed specialist agents - Skill authors and maintainers who want scenario-tested, reviewable skills - Game, engineering, frontend, support, product, marketing, and research workflows - Creator, influencer, and founder-led content workflows ## What is included - Atomic skills under `skills/` - Installable skillsets under `skillsets/` - MCP presets for tool-aware workflows - `AGENTS.md` routing templates - Host starter templates under `templates/` - Host-specific setup adapters and docs - Validation scripts - Scenario tests - Evaluation templates for before/after reports, quality rubrics, and security reviews - Documentation and examples for setup, use, and skillset authoring ## Validation Status Current v0.2 development gate: - 166 validated skill files - 33 validated skillsets - 123 routing scenarios - all skillset install dry-runs pass - fresh Codex home install smoke passes for `game-dev` and `all` - `gh skill publish --dry-run` returns `ok` ## Example use cases Use AgentSkills when you want an AI coding agent to behave more like a specialist: - Onboard AgentSkills into a new repo with focused project routing - Build and QA an HTML5 game - Prepare a weekly executive brief and decision memo - Draft investor updates and fundraising readiness artifacts - Map AI use cases and pilot governance before rollout - Create a small-business owner dashboard and cash action list - Run a weekly leadership cadence with decisions, risks, and actions - Turn engineering issues into scoped plans and release checks - Plan and implement a Three.js interactive page - Prepare a product launch workflow - Synthesize product feedback into a reviewed PRD - Use a guided "grill me" interview to fill requirement gaps before drafting - Turn stakeholder notes into a BRD with traceable business requirements - Triage support tickets and draft reviewed customer replies - Draft KPI narratives and variance commentary for review - Run structured competitive research - Create a repeatable frontend/product implementation workflow - Author new reusable agent skills - Benchmark, score, and improve skills with before/after reports - Audit skills before installing or publishing them - Improve READMEs, API docs, release notes, and migration guides - Design backend APIs, schemas, auth flows, and service boundaries - Plan deployments, CI workflows, rollback, and production readiness - Build focused test plans, QA matrices, bug reproductions, and release gates - Turn analytics questions into metric definitions, dashboards, and read-only analysis plans - Build creator content pillars, hooks, scripts, calendars, captions, and repurposing plans - Prepare creator media kits, sponsor pitches, campaign proposals, rate cards, and recap reports - Turn creator analytics exports into sponsor reports, content learnings, affiliate reports, and dashboard specs - Map creator monetization options into offer ladders, launch plans, pricing reviews, and owned-revenue dashboards - Run creator business ops reviews for sponsor pipeline, invoice readiness, production handoffs, and asset systems - Triage creator comments and DMs into reviewed replies, moderation briefs, feedback themes, polls, and engagement plans - Review creator reputation risks for disclosures, rights, AI labeling, brand safety, platform policy, and public responses - Plan AI-assisted creator production with prompt packs, synthetic character bibles, provenance notes, rights checks, and human-review gates ## Repository structure ```text AgentSkills/ |-- skills/ |-- skillsets/ |-- mcp/ |-- agents/ |-- templates/ |-- scripts/ |-- tests/ `-- docs/ ``` ## Getting started 1. Clone the repository. ```powershell git clone https://github.com/jeremylongworth-source/AgentSkills.git cd AgentSkills ``` 2. Install a skillset. ```powershell .\scripts\install-skillset.ps1 game-dev ``` 3. Review the installed instructions before using them with your agent. 4. Start your coding agent with the relevant skillset active. ## Host setup guides Use the host-specific guides when installing AgentSkills into a particular coding agent: - [Codex](docs/setup/codex.md) - [GitHub Copilot](docs/setup/github-copilot.md) - [Cursor](docs/setup/cursor.md) - [Claude Code](docs/setup/claude-code.md) - [Gemini CLI](docs/setup/gemini-cli.md) Start with [setup guides](docs/setup/README.md) if you are choosing between `gh skill install`, local skillset scripts, project instruction files, and MCP presets. See [examples](docs/examples/README.md) for single-skill installs, skillset installs, MCP preset usage, and before/after agent behavior. See [compatibility](docs/compatibility.md) for host paths and MCP formats, and [troubleshooting](docs/troubleshooting.md) for common setup failures. For new multi-skill bundles, see the [bundle authoring guide](docs/bundle-authoring-guide.md). For the first expanded business bundle, see the [support success brief](docs/bundles/support-success.md). For skill and bundle evaluation workflows, see the [agentops evaluation brief](docs/bundles/agentops-evaluation.md). For business analysis and BRD workflows, see the [business analysis brief](docs/bundles/business-analysis.md). For skill install and publish review workflows, see the [skill security audit brief](docs/bundles/skill-security-audit.md). For documentation workflows, see the [technical documentation brief](docs/bundles/technical-documentation.md). For backend and API design workflows, see the [backend API brief](docs/bundles/backend-api.md). For DevOps, cloud, and release workflows, see the [DevOps cloud release brief](docs/bundles/devops-cloud-release.md). For quality and testing workflows, see the [quality testing brief](docs/bundles/quality-testing.md). For data analytics and BI workflows, see the [data analytics BI brief](docs/bundles/data-analytics-bi.md). For creator content workflows, see the [creator content engine brief](docs/bundles/creator-content-engine.md). For creator sponsorship workflows, see the [creator brand deals brief](docs/bundles/creator-brand-deals.md). For creator analytics and reporting workflows, see the [creator analytics reporting brief](docs/bundles/creator-analytics-reporting.md). For creator monetization workflows, see the [creator monetization brief](docs/bundles/creator-monetization.md). For creator business operations workflows, see the [creator business ops brief](docs/bundles/creator-business-ops.md). For creator community workflows, see the [creator community brief](docs/bundles/creator-community.md). For creator reputation risk workflows, see the [creator reputation risk brief](docs/bundles/creator-reputation-risk.md). For creator AI production workflows, see the [creator AI production brief](docs/bundles/creator-ai-production.md). For portability rules, see [vendor neutrality](docs/vendor-neutrality.md). For expansion planning, see the [scope broadening roadmap](docs/scope-broadening-roadmap.md). For mapping new bundle proposals, see the [bundle expansion map](docs/bundle-expansion-map.md). For current scope decisions, see [bundle expansion decisions](docs/bundle-expansion-decisions.md). For the executive bundle expansion, see [executive skillsets roadmap](docs/executive-skillsets-roadmap.md). For the creator and influencer expansion, see [creator skillsets roadmap](docs/creator-skillsets-roadmap.md). For the active research and development backlog, see the [roadmap research plan](docs/roadmap-research-plan.md). For skill proof, before/after checks, and security review templates, see [evaluation](docs/evaluation/README.md). For v0.2 promotion checks, see the [alpha validation plan](docs/evaluation/v0.2-alpha-validation-plan.md). For validation status, see the [v0.2 validation tracker](docs/evaluation/v0.2-validation-tracker.md). For v0.2 release-candidate scope and prerelease decision support, see [v0.2 release candidate checklist](docs/release-candidate-v0.2.md). For maintainer handoff before tagging, see the [v0.2 release handoff](docs/v0.2-release-handoff.md). For the path from prerelease to normal release, see the [stable release plan](docs/stable-release-plan.md). For release details, see [v0.2.2 release notes](docs/release-notes/v0.2.2.md). ## Suggested first skillset For a new repository where you want focused AgentSkills setup, start with the `agentskills-project-onboarding` skill. It should inspect the repo, recommend a minimal skillset or small skill list, draft project routing guidance, and give verification steps before install commands are run. For game development or interactive browser work, start with: ```powershell .\scripts\install-skillset.ps1 game-dev ``` For browser game publishing workflows, use: ```powershell .\scripts\install-skillset.ps1 html5-game-publishing ``` ## Supported agent workflows AgentSkills is intended for workflows involving: - OpenAI Codex - GitHub Copilot - Cursor - Claude Code - Gemini CLI - MCP-enabled agent systems - Other coding agents that can read project instructions and reusable skill files Exact host support may vary depending on how each tool loads instructions, skills, and workspace context. ## Roadmap Planned improvements include: - Forward-testing for AgentOps evaluation and skill security audit bundles - More forward-test reports and expanded vertical bundles - Additional executive, owner-operator, customer success, and vertical bundles - Creator platform expansion, live/event, and additional vertical bundles - More scenario tests, before/after reports, and security review examples - More MCP presets and host setup examples where they improve portability ## Contributing Contributions are welcome. Good first contributions include: - Adding setup instructions for your preferred coding agent - Improving install scripts - Adding examples - Adding tests - Proposing new skillsets - Improving existing skill instructions See [Contributing](CONTRIBUTING.md), [Changelog](CHANGELOG.md), [Security](SECURITY.md), and [Code of Conduct](CODE_OF_CONDUCT.md) for contribution, release, vulnerability reporting, and community guidance. ## License MIT ## Star the repo If AgentSkills helps your AI development workflow, star the repo so other builders can find it.