The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill and its associated check-upstream.sh script fetch documentation, integration guides, and package metadata from Convex's official GitHub repositories and documentation domain (docs.convex.dev). These downloads are used to ensure the integration remains compatible with the latest upstream standards.
[COMMAND_EXECUTION]: The skill executes a local shell script (scripts/check-upstream.sh) to perform update checks and uses npm to install dependencies such as @convex-dev/self-hosting. It also provides instructions for running deployment tools via npx.
[PROMPT_INJECTION]: The skill exposes an indirect prompt injection surface by instructing the agent to read and follow implementation rules from external, remote files like INTEGRATION.md and llms.txt.
Ingestion points: External content enters the agent context via remote documentation URLs and the output of the check-upstream.sh script.
Boundary markers: The instructions do not define delimiters or provide warnings to the agent to treat remote content as untrusted data.
Capability inventory: The skill possesses the capability to modify project files (e.g., convex/convex.config.ts) and execute shell commands (npm install, npx).
Sanitization: No validation or sanitization process is implemented for the content fetched from remote markdown or text files before the agent acts upon it.

convex-self-hosting