The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The maintenance script scripts/check-updates.py fetches version metadata from the official npm registry (registry.npmjs.org) to verify if the skill is up to date with the upstream software. This is a legitimate diagnostic operation targeting a well-known service.
[COMMAND_EXECUTION]: The documentation includes standard development commands for the Convex platform, such as npm install, npx convex dev, and npx convex deploy. These instructions are intended for setting up and managing backend infrastructure and do not represent malicious command execution.
[INDIRECT_PROMPT_INJECTION]: The skill provides patterns for building applications that handle external data, such as webhooks and AI chat interfaces. It explicitly documents and mitigates potential injection risks by emphasizing the use of the platform's built-in schema validators and authentication checks.
Ingestion points: Described in references/02-functions-actions-http.md and references/10-ai-agents.md for processing external data through HTTP endpoints and LLM conversation history.
Boundary markers: The documentation requires the use of strict schema validators (v library) for all database operations, which serve as data boundaries.
Capability inventory: The functions documented include database writes, scheduling, and external API requests (Actions).
Sanitization: The skill promotes the use of typed validators and server-side authentication verification to ensure all external data is sanitized and authorized.

convex