Menu
Add or extend a Repo Dungeon GitHub data flow, including Octokit access, typed models, caching, backoff, and mocked verification coverage.
Installer avec Codex ou Claude Copiez ce prompt, collez-le dans Codex, Claude ou un autre assistant, puis laissez-le vérifier la page du skill et l'installer pour vous.
Basé sur la classification professionnelle SOC
| name | implement-github-data-flow |
| description | Add or extend a Repo Dungeon GitHub data flow, including Octokit access, typed models, caching, backoff, and mocked verification coverage. |
Use this skill when Repo Dungeon needs a new GitHub API call, repo-detail payload, or authentication-related data path.
Start from the PRD's GitHub API section and identify the exact REST endpoint, scope requirements, pagination rules, and fallback behavior.
Update the typed models in src/github/types.ts so downstream systems receive normalized data rather than raw API payloads.
Add or update the Octokit call, caching policy, retry behavior, and error normalization in src/github/api.ts or src/game/systems/RoomLoader.ts.
export async function fetchExample(): Promise<ExampleResult> {
// request
// normalize
// cache
// return typed payload
}
Surface the new capability through a hook or typed helper, ensuring tokens remain client-only and README-like content stays plain text if rendered.
Test unauthenticated limits, authenticated success, 404/403/429 handling, truncation cases, and mocked success paths before handing the flow to UI or gameplay agents.
See docs/PRD.md for the full specification:
Add a complete Repo Dungeon biome content pack, including tileset references, prop/NPC dressing, palette cues, and ambient audio hooks for a language or topic zone.
Add Repo Dungeon unit, component, or integration coverage for a gameplay, UI, or platform subsystem using Vitest, React Testing Library, and MSW as appropriate.
Scaffold a Repo Dungeon Phaser scene with lifecycle hooks, typed event wiring, and integration points for generation, UI overlays, and progression events.
Build an accessible Repo Dungeon React overlay, panel, or screen that consumes shared game state without moving feature logic out of its specialist owner.
Discover which LLM models are available to the user (cloud subscription + local Ollama), classify each generated agent's workload, and recommend (and optionally apply) a per-agent model assignment so that lightweight agents do not default to the most expensive model. Use this skill after `forge-build-agent-team` has produced an agent team, or any time the team changes.
Meta-skill that bootstraps a new project end-to-end from a one-liner idea by chaining the deterministic Forge skills: forge-build-prd → (pause for review) → forge-build-agent-team → (pause for review) → optionally forge-assign-models. Use this skill when a user wants to go from idea to a reviewable PRD and a generated agent team in a single guided flow, without losing the human review gates between steps.