The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill provides patterns and code snippets for asynchronous web scraping and concurrent I/O processing (e.g., scrape_urls and fetch_pages in resources/implementation-playbook.md). These patterns involve ingesting untrusted remote content, which is a known attack surface for indirect prompt injection if the agent subsequently processes that data without sanitization.
Ingestion points: fetch_url and scrape_urls functions in resources/implementation-playbook.md retrieve content from remote URLs.
Boundary markers: The provided educational snippets do not demonstrate the use of prompt boundaries or explicit instructions to ignore embedded directives in the fetched content.
Capability inventory: The skill patterns utilize network access via aiohttp.ClientSession and simulated database interactions.
Sanitization: No data validation or sanitization logic is included in the high-level educational examples for handling remote data.
[EXTERNAL_DOWNLOADS]: The resource documentation in resources/implementation-playbook.md references several well-known and trusted Python libraries, including aiohttp, pytest-asyncio, FastAPI, asyncpg, and motor. These are standard tools in the Python ecosystem for asynchronous development and do not represent a security risk in this context.

async-python-patterns