The Agent Skills Directory

[DATA_EXPOSURE_AND_EXFILTRATION]: The skill implements a dedicated security scanning phase that identifies tracked secrets, PII, and private environment files (e.g., .env, .pem, .key) using grep patterns, ensuring they are not accidentally included in public repository releases.\n- [COMMAND_EXECUTION]: Utilizes standard developer tools including git, rsync, and the GitHub CLI (gh) to manage project visibility, repository tagging, and marketplace metadata distribution.\n- [EXTERNAL_DOWNLOADS]: Installs the mcp-publisher utility via the Homebrew package manager, which is a well-known service, to facilitate server registration with the Model Context Protocol registry.\n- [PERSISTENCE_MECHANISMS]: Recommends configuring core.hooksPath to point to a repository-controlled directory (.githooks) to automate the synchronization of skill copies during git events, which is a standard practice for maintainable project automation.\n- [INDIRECT_PROMPT_INJECTION]: The skill processes repository files via git ls-files and Read tools for validation and auditing (ingestion points). While it uses grep-based patterns for filtering (sanitization) rather than explicit boundary markers, its capabilities (Bash, Write) are applied in a low-risk context focused on static verification of the user's own codebase.

publish-skills