The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill fetches an OpenAPI document from a local endpoint, which is then used to influence agent summaries and downstream code generation.
Ingestion points: Data is retrieved from https://localhost:44339/umbraco/openapi/management.json in SKILL.md.
Boundary markers: The instructions do not define delimiters or warnings to ignore potentially malicious instructions embedded within the fetched JSON data.
Capability inventory: The skill has the capability to write to the file system (OpenApi.json), execute build commands via dotnet run, and run generation scripts via npm run (referenced in SKILL.md).
Sanitization: While the skill mentions that the generate:openapi script ensures the response is valid JSON before writing, there is no verification or sanitization of the content itself for prompt injection attacks.
[DYNAMIC_EXECUTION]: The skill uses the npm run generate:server-api command to regenerate a TypeScript client (hey-api) based on the content of the OpenApi.json file. This represents the runtime generation of executable source code based on data provided by an external, albeit local, server process.
[COMMAND_EXECUTION]: The skill utilizes several local shell commands to manage the development environment:
dotnet run is used to start the backend application service.
npm run is used to trigger project-specific scripts for fetching data and generating code.
curl is used to probe the status of the local API endpoint.

umb-update-openapi