The Agent Skills Directory

EXTERNAL_DOWNLOADS (MEDIUM): The skill's primary function is to download and install external packages from various sources (GitHub, etc.) via the npx skills add command. While the documentation mentions trusted sources like vercel-labs/agent-skills, it allows for the installation of any package found via keyword search.
REMOTE_CODE_EXECUTION (MEDIUM): The skill promotes the use of npx, a tool that executes code from the npm registry or remote repositories. It specifically suggests the -y flag (e.g., npx skills add <package> -g -y) to skip confirmation prompts, which could lead to the unintended execution of malicious code if a user is tricked into installing a compromised skill.
COMMAND_EXECUTION (MEDIUM): User-provided search queries are directly interpolated into shell commands (e.g., npx skills find [query]). This creates a surface for command injection if the underlying execution environment does not strictly sanitize the input string.
INDIRECT_PROMPT_INJECTION (LOW): This skill exhibits a vulnerability surface to indirect prompt injection.
Ingestion points: Untrusted data enters via search queries and the textual output of the npx skills find command (SKILL.md).
Boundary markers: None present in the instructions to separate search results from system instructions.
Capability inventory: The skill utilizes npx for command execution and package installation.
Sanitization: No evidence of input sanitization or validation of the search results before presenting them to the user or executing installation commands.

find-skills