The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes the npx skills command to perform searches and manage package installations.
[REMOTE_CODE_EXECUTION]: The skill recommends the command npx skills add <package> -g -y to install new capabilities. This instruction facilitates the global installation of code from remote GitHub repositories. The use of the -y flag is a significant risk as it bypasses user confirmation prompts, potentially leading to the unattended execution of malicious code from third-party sources.
[EXTERNAL_DOWNLOADS]: The skill connects to the external registry at https://skills.sh/ to search for skills and downloads content from various GitHub repositories during the installation process.
[PROMPT_INJECTION]: The skill's workflow creates an attack surface for indirect prompt injection via the external registry.
Ingestion points: Results and metadata from npx skills find are retrieved from https://skills.sh/ and presented to the agent in SKILL.md.
Boundary markers: No delimiters or "ignore instructions" warnings are used to isolate content fetched from the external registry.
Capability inventory: The skill has the capability to execute shell commands and install software packages via npx.
Sanitization: There is no evidence of sanitization or validation of the data retrieved from the registry before it is processed by the agent.

find-skills