The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill provides patterns for building applications that consume data from external sources (e.g., APIs and data streams), creating a potential attack surface if the processed data is untrusted.
Ingestion points: Coroutines like fetch_user, fetch_page, and fetch_data are designed to process external inputs.
Boundary markers: No explicit delimiters or boundary markers for untrusted data are used in the prompt interpolation examples.
Capability inventory: The skill examples utilize network operations via aiohttp and requests, as well as file system access via fs.readFile and fs/promises.
Sanitization: The provided patterns do not demonstrate explicit sanitization or validation of input data before processing.
[COMMAND_EXECUTION]: The workflow section suggests the use of shell commands such as pytest and grep for code verification and environment auditing. These represent standard developer tool invocations within the local development environment.

async-expert