The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill requires the installation of the mcp-publisher CLI tool using the Homebrew package manager to enable publishing to MCP registries.- [COMMAND_EXECUTION]: Instructions involve executing shell commands for software installation and tool-specific operations for authentication and publishing (e.g., brew install, mcp-publisher login, mcp-publisher publish).- [INDIRECT_PROMPT_INJECTION]: The skill guides an agent to interact with a wide array of external directory sites, registries, and GitHub repositories, creating an attack surface where malicious content on those sites could influence agent behavior.
Ingestion points: Form fields, site metadata, and Pull Request templates from 50+ external directory domains.
Boundary markers: The skill does not define specific delimiters for data ingested via browser automation, relying instead on high-level operational guidelines.
Capability inventory: Capabilities include writing to the local filesystem (manifests/logs), browser automation, and performing GitHub Pull Request operations.
Sanitization: The skill includes 'Hard Operational Rules' that mitigate risks by filtering for honeypots and enforcing factual, non-AI-generated copy styles.

directory-submissions