The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes git clone --depth 1 using a repository URL provided by the user to download third-party skill directories to a scratch location.
[COMMAND_EXECUTION]: The skill runs npx -y skills add . -l, which executes a Node.js package to list and verify the state of the local skills directory, including the newly imported content.
[DATA_EXFILTRATION]: The skill performs recursive file deletion using find . -name skills-lock.json -delete. While intended to maintain repository standards by removing lock files, it represents a file system modification action triggered by external content.
[REMOTE_CODE_EXECUTION]: By design, the skill fetches code and documentation from arbitrary external GitHub repositories and vendors them into the local environment. This facilitates the introduction of third-party logic into the agent's available toolkit.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection vulnerabilities.
Ingestion points: External skill directories (containing SKILL.md and supporting files) are fetched from user-specified GitHub URLs.
Boundary markers: No boundary markers or isolation techniques are applied to the imported content.
Capability inventory: The skill possesses capabilities to write to the file system, delete files, and execute commands via git and npx.
Sanitization: No sanitization or validation of the downloaded instructions is performed before they are placed into the local catalog and processed by the verification tool.

import-skill