The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides instructions for the agent to register a new MCP server using command-line tools such as codex mcp add or claude mcp add. These instructions target https://developers.openai.com/mcp, which is an official domain belonging to a well-known service.
[INDIRECT_PROMPT_INJECTION]: The skill has an attack surface for indirect injection as it retrieves and processes content from external documentation searches.
Ingestion points: Documentation content is retrieved from external sources via the mcp__openaiDeveloperDocs__search_openai_docs and mcp__openaiDeveloperDocs__fetch_openai_doc tools specified in SKILL.md.
Boundary markers: The instructions require the agent to provide citations and prefer paraphrasing, though no specific technical delimiters (like XML tags or clear boundaries) are defined for the fetched documentation strings.
Capability inventory: The skill is limited to searching and fetching documentation; it does not include scripts with file-write, arbitrary command execution, or outbound network capabilities beyond the MCP tool transport.
Sanitization: No explicit sanitization or filtering of the documentation content is defined in the instructions.

openai-docs