The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill connects to https://skills.sh/ to retrieve metadata and can download package content from various GitHub repositories.
[REMOTE_CODE_EXECUTION]: Uses the npx skills add command to install and execute external skill modules. The instructions advocate for using the -y flag, which allows for installation and execution without interactive user confirmation.
[COMMAND_EXECUTION]: Executes shell-based CLI commands (npx skills find, add, check, update) to manage the local agent environment.
[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface (Category 8) by processing untrusted metadata from a public registry.
Ingestion points: Metadata and descriptions returned by the npx skills find command (SKILL.md).
Boundary markers: Absent; there are no clear delimiters or instructions for the agent to ignore potentially malicious content within skill descriptions.
Capability inventory: The skill has the capability to modify the agent's environment by adding or updating executable code via npx skills (SKILL.md).
Sanitization: There is no evidence of sanitization or validation of the text retrieved from the external registry before it is parsed by the agent.

find-skills