The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill's primary function is to audit and process external project data, specifically the convex/ directory and application schema. This creates a surface for indirect prompt injection where malicious content embedded in the audited files could attempt to manipulate the agent's assessment or the resulting optimization plan.
Ingestion points: Application files within the convex/ directory and deployment schema.
Boundary markers: The instructions do not specify any delimiters or safety prompts to ensure the agent treats project content strictly as data.
Capability inventory: The skill has the capability to modify project files and execute management tools (upgrades and observability installation).
Sanitization: There is no mention of sanitizing or filtering the content of the files being audited before they are processed by the agent.
[COMMAND_EXECUTION]: The workflow involves running assessment and update tools such as launch-readiness and check-updates, and offers to install additional components like sentinel. While these appear to be part of the Convex development ecosystem, they represent the execution of logic based on the state of external project dependencies.

convex-optimize