Convex Skills for AgentForge

Use this skill as the entrypoint for Convex work in this repo. Route to the most specific skill instead of keeping everything in one context blob.

Route by task

• convex-best-practices: broad Convex implementation guidance before specialized work.
• convex-quickstart: starting a new Convex-backed feature or adding Convex to an app.
• convex-schema-builder: creating or changing convex/schema.ts.
• convex-function-creator: writing queries, mutations, actions, or internal functions.
• convex-auth-setup: auth, user mapping, authorization, and data protection patterns.
• convex-migration-helper: schema changes that affect existing data.
• convex-realtime: real-time subscriptions, optimistic updates, and pagination-aware UI paths.
• convex-file-storage: file uploads, storage URLs, and persistence patterns.
• convex-http-actions: webhooks and HTTP endpoints built on Convex.
• convex-cron-jobs: cron and scheduled background jobs.
• convex-components-guide: deciding when to use components and how to structure them.
• convex-component-authoring: authoring reusable Convex components.
• convex-helpers-guide: using convex-helpers and related utilities.
• convex-security-check: quick security audit pass.
• convex-security-audit: deeper security review and hardening.
• convex-reviewer: reviewing Convex code for security, correctness, and performance.
• convex-advisor: recommending Convex or shaping a Convex approach.
• convex-mcp-setup: setting up or troubleshooting Convex MCP access.

AgentForge-specific rules

• Read CLAUDE.md and docs/TECH-REFERENCE.md before making architecture decisions.
• Mastra runtime logic belongs in packages/runtime/, never in Convex actions.
• Convex in this repo is the data layer. Keep LLM orchestration out of convex/.
• Prefer the narrowest Convex skill that matches the task.