菜单
Drop and recreate the local dev Postgres database, then reapply every bounded context's EF Core migrations in dependency order. Refuses to run against anything but localhost. Invoke when the user says "reset dev db", "reset local database", "wipe my local postgres", or "/reset-dev-db".
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
Cross-reference Application write-service methods / Angular components against existing test files, ranked by how central each file is (referenced-by count), to surface the highest-leverage places to add tests next. Invoke when the user says "find untested code", "what's untested", "test coverage gaps", or "/find-untested".
Scaffold a brand-new bounded context (all backend layer projects, solution wiring, API gateway route, and the matching Angular feature folder). Invoke when the user says "new bounded context", "scaffold context", "add a bounded context", or "/new-bounded-context <ContextName>".
Brainstorm at least 10 automation ideas for this project across three tiers (plain non-AI automation, AI-assisted SDLC support, AI agentic loops), then create one GitHub issue per idea the user picks. Invoke when the user says "automation ideas", "how could we automate this project", "suggest some automations", or "/automation-ideas".
Report outdated and vulnerable packages across the backend (Directory.Packages.props / Hiscary.sln) and frontend (client/package.json) in one combined summary. Invoke when the user says "dependency check", "check for outdated packages", "any vulnerable dependencies", or "/dependency-check".
Implement a spec generated by /spec. Reads .claude/specs/{slug}/design.md, executes each task in Section 8 in order, marks tasks complete as it goes, verifies acceptance after each task, and stops on failure. Invoke when the user says "implement issue
Report open GitHub issues/PRs, failing CI runs, pending EF Core migrations per bounded context, Nx-affected projects since the last release tag, stale branches, and outdated/vulnerable packages — one combined health snapshot. Invoke when the user says "project health", "health check", "how healthy is the repo", or "/project-health".
| name | reset-dev-db |
| description | Drop and recreate the local dev Postgres database, then reapply every bounded context's EF Core migrations in dependency order. Refuses to run against anything but localhost. Invoke when the user says "reset dev db", "reset local database", "wipe my local postgres", or "/reset-dev-db". |
| allowed-tools | Bash Read |
/reset-dev-db [--seed]
This skill must refuse to run against anything but the local dev connection string. Before doing anything destructive, resolve the connection string the same way Hiscary.AppHost does (via dotnet user-secrets list --project server/src/Hiscary.AppHost, key Parameters:postgres-password, host always localhost:5432 per server/CLAUDE.md) and verify the host is literally localhost or 127.0.0.1. If it resolves to anything else — a remote host, an .env-provided override pointing elsewhere, or the secret is missing entirely — stop and report the resolved host without executing any drop/recreate command. Never accept a connection string from a flag or environment variable without this check; a typo'd POSTGRES_HOST pointing at a shared or prod-like environment must not be able to bypass the guard.
dotnet user-secrets list --project server/src/Hiscary.AppHost | grep postgres-password
Print the resolved host (should be localhost:5432) and ask the user to confirm before proceeding if this is the first time the skill has been run in a session (subsequent invocations in the same session don't need to re-confirm).
The local Postgres container is Aspire-managed, so the cleanest reset is via psql against the running container rather than restarting the container itself (restarting would also reset volumes for other Aspire-managed services unnecessarily):
docker exec -it $(docker ps --filter "name=hiscary-" --format "{{.Names}}" | head -1) \
psql -U postgres -c "DROP DATABASE IF EXISTS postgres;" -c "CREATE DATABASE postgres;"
If no hiscary-* container is running, tell the user to start the backend first (./scripts/start-backend.sh) — this skill resets an already-running local Postgres, it doesn't start Aspire itself.
Run dotnet ef database update per context. Order matters where contexts reference shared schema setup — run UserAccounts and PlatformUsers first (identity/user data other contexts' seed data may assume exists), then the rest:
dotnet ef database update --project server/src/Hiscary.UserAccounts.Persistence.Context
dotnet ef database update --project server/src/Hiscary.PlatformUsers.Persistence.Context
dotnet ef database update --project server/src/Hiscary.Stories.Persistence.Context
dotnet ef database update --project server/src/Hiscary.Notifications.Persistence.Context
Media and Recommendations currently have no Persistence.Context project (no EF-backed schema as of this writing) — skip them but say so explicitly rather than silently omitting two of the six contexts from the report.
If --seed was passed, run any existing *Seeder classes (e.g. Hiscary.Stories.Persistence.Context/Seeds/GenreSeeder.cs) that are already wired into OnModelCreating/startup — EF's HasData seeds apply automatically on database update, so this step is usually a no-op unless a context has a separate manual seeding entry point. Check before assuming one exists.
Confirm all four EF-backed contexts are at their latest migration (dotnet ef migrations list --project ... | tail -1 should match the newest migration file) and note the two schema-less contexts. If any database update fails, stop and report the exact error — don't proceed to the next context with a half-applied schema.