The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill instructs the agent on how to fetch and process data from external sources such as databases, caches, and the file system. These operations represent a vulnerability surface where maliciously crafted data could attempt to influence the agent's behavior.
Ingestion points: Database query results in examples/core.md, cache retrieval in examples/caching.md, and file reading in examples/async.md.
Boundary markers: The skill does not provide specific delimiters or instructions to the agent to ignore potentially malicious content within the fetched data.
Capability inventory: The skill allows for database read/write/delete operations, cache management, and file reading.
Sanitization: While the skill correctly promotes parameterized SQL queries to prevent SQL injection, it does not address sanitization for prompt injection within the data context.
[DYNAMIC_EXECUTION]: The skill provides implementation patterns for Node.js worker_threads to handle CPU-bound tasks. This involves creating new execution contexts at runtime.
Evidence: The WorkerPool class in examples/async.md uses the Worker constructor to execute a script specified by a relative path (./workers/data-processor.worker.js) with dynamic workerData provided at runtime.

api-performance-api-performance