The Agent Skills Directory

[SAFE]: The skill logic and instructions align with its stated purpose of web research and monitoring. It follows security best practices by recommending environment variables for API keys and using temporary directories for local caches.
[COMMAND_EXECUTION]: The skill utilizes the Bash tool for legitimate data processing and retrieval tasks, such as using jq for JSON manipulation, diff for comparing snapshots, and curl for interacting with the Tavily API.
[EXTERNAL_DOWNLOADS]: Communicates with the official Tavily API (api.tavily.com) for semantic search and batch content extraction. This is a well-known service for research automation.
[PROMPT_INJECTION]: The skill possesses an inherent attack surface for indirect prompt injection as it ingests untrusted content from external websites.
Ingestion points: Retrieves external web content via WebFetch, Tavily, and agent-browser (e.g., in rules/browser-patterns.md).
Boundary markers: The prompts do not specify clear delimiters or isolation instructions to the agent when processing the retrieved content.
Capability inventory: The agent has access to Bash (command execution) and Write (filesystem access), which are powerful capabilities if influenced by malicious instructions in scraped data.
Sanitization: No explicit sanitization or filtering of the extracted web content is implemented before it enters the model's context.

web-research-workflow