The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill uses npx --yes skills to download and execute the skills CLI from Vercel Labs' public repository. This is used for adding local repositories for testing and installing published skills to the agent's environment.
[COMMAND_EXECUTION]: The skill executes various shell commands via the skills CLI, including skills add --list, skills add --skill <name>, and skills add <owner>/<repo>. These commands are used to discover, test, and perform global installations of agent skills.
[INDIRECT_PROMPT_INJECTION]: The skill processes external data by reading SKILL.md files and entire GitHub repositories to package and validate them. This ingestion surface is susceptible to indirect prompt injection if the source content contains malicious instructions.
Ingestion points: Reads files in skills/<name>/SKILL.md, target repositories, and remote GitHub tree URLs.
Boundary markers: The skill instructions include a validation phase to ensure that instructions carry only authorized work through stated boundaries.
Capability inventory: Uses npx for CLI execution and git for repository management (commit, PR creation, diff inspection).
Sanitization: Explicitly instructs the agent to inspect the publication diff for local paths, logs, credentials, and private examples before release.

skill-maker