The Agent Skills Directory

[SAFE]: The skill provides educational content and code snippets for software development. No malicious behavior, obfuscation, or unauthorized access patterns were detected.
[EXTERNAL_DOWNLOADS]: The skill includes code snippets referencing the official @anthropic-ai/sdk. This is a well-known package from a major service provider and is documented as a standard dependency for the intended functionality.
[INDIRECT_PROMPT_INJECTION]: The skill acknowledges the risk of processing external data by including dedicated stages for 'Input Validation + Sanitization' and 'Output Parsing + Validation' in its recommended architecture. These are standard defensive measures for applications that interact with LLMs.
Ingestion points: User-provided userInput in the generateContent function.
Boundary markers: The architectural diagram explicitly includes validation and sanitization steps before prompt construction.
Capability inventory: The code demonstrates network requests to the Anthropic API and interactions with a local database for usage tracking (db.usage.create).
Sanitization: The code includes an example of input length validation and suggests further sanitization steps.

ai-wrapper-product