The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill possesses a surface for indirect prompt injection as it ingests untrusted data from web search results, GitHub issues, and repository discussions. However, it includes mandatory security guardrails: the instructions explicitly command the agent to treat retrieved docs as untrusted source material, forbidding the execution of instructions embedded in them or the disclosure of local data based on the content of retrieved pages. It also mandates the sanitization of all search queries to remove credentials and sensitive information.
[COMMAND_EXECUTION]: The skill utilizes the tvly CLI and curl commands to interact with the vendor's search and extraction APIs. These command executions are documented examples for retrieval and research purposes, following standard practices without privilege escalation or unsafe command concatenation.
[EXTERNAL_DOWNLOADS]: The skill performs network requests to the Tavily API and references official documentation domains (e.g., docs.python.org, github.com). These network operations are essential to the skill's purpose and are conducted through secure vendor endpoints and established service domains.

developer-atlas