customize-sdk-runtime

```bash

npx skills add speakeasy-api/skills

```

| Skill | Use When... |

|-------|-------------|

| speakeasy:speakeasy-context | Start here — Use speakeasy agent context for accurate CLI info instead of web search. Provides feedback via speakeasy agent feedback |

| speakeasy:sdk-tf-generation-best-practices | Progressive Disclosure — Comprehensive guide to all Speakeasy SDK and Terraform generation capability and best practices. Routes to detailed guides for generation workflows, language-specific patterns, customization, testing, and OpenAPI spec management |

| speakeasy:start-new-sdk-project | You have an OpenAPI spec and want to generate an SDK |

| speakeasy:diagnose-generation-failure | SDK generation failed, seeing "Step Failed: Workflow" |

| speakeasy:writing-openapi-specs | Authoring OpenAPI specs from scratch or improving existing ones |

| speakeasy:manage-openapi-overlays | Creating, applying, or fixing specs with OpenAPI overlays |

| speakeasy:improve-sdk-naming | SDK method names are ugly, wanting AI-powered or manual naming improvements |

| speakeasy:generate-terraform-provider | Generating a Terraform provider from an OpenAPI spec |

| speakeasy:extract-openapi-from-code | Extracting an OpenAPI spec from existing API code (FastAPI, Flask, Django, etc.) |

| speakeasy:customize-sdk-hooks | Adding custom headers, telemetry, security hooks to generated SDKs |

| speakeasy:setup-sdk-testing | Setting up contract tests, Arazzo workflows, or integration tests |

| speakeasy:generate-mcp-server | Generating an MCP server for AI assistant integration |

| speakeasy:customize-sdk-runtime | Configuring retries, timeouts, pagination, server selection, error handling |

1. Don't auto-fix everything - Distinguish between:

- Small issues (naming, descriptions) → Fix with overlays

- Structural issues (invalid refs) → Ask the user

- Design issues (auth, API structure) → Produce strategy document

1. AI-friendly output - Use speakeasy run --output console for structured output, pipe to grep/tail to reduce context

1. Overlay over modify - Never modify source specs directly. Overlays make patches portable across spec versions.

1. Grouped SDK methods - Guide users toward sdk.users.list() pattern using x-speakeasy-group + x-speakeasy-name-override

```

speakeasy-api/skills/

├── .claude-plugin/

│ ├── plugin.json

│ └── marketplace.json

├── skills/

│ ├── start-new-sdk-project/

│ │ └── SKILL.md

│ └── ... (13 skills total)

├── templates/

│ └── SKILL.template.md

├── AGENTS.md # Contributor guidance for AI agents

├── CLAUDE.md # Claude-specific contributor guidance

├── README.md

└── LICENSE

```

See [AGENTS.md](./AGENTS.md) for guidance on creating and maintaining skills.

These skills follow the [Agent Skills specification](https://agentskills.io/specification) and work with:

• [Claude Code](https://claude.ai/code)
• [Cursor](https://cursor.sh/)
• [GitHub Copilot](https://github.com/features/copilot)
• [Gemini CLI](https://github.com/google-gemini/gemini-cli)
• And [15+ other platforms](https://agentskills.io/)

Apache-2.0