The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill instructs the agent to download and install code from external GitHub repositories via the npx skills add command. It explicitly suggests using the -y flag to skip confirmation prompts, which allows for the silent installation and execution of arbitrary code from third-party sources.
[COMMAND_EXECUTION]: The skill relies on shell commands (npx skills ...) for its core functionality. It is vulnerable to command injection if a user-supplied query or package name (e.g., from a malicious task description) is interpolated into the shell command without proper escaping.
[EXTERNAL_DOWNLOADS]: The skill facilitates downloads from https://skills.sh and arbitrary GitHub repositories. While intended for skill discovery, this mechanism allows for the retrieval of unverified content from untrusted external sources.
[PROMPT_INJECTION]: The skill exhibits a surface for indirect prompt injection where malicious instructions could be embedded in the search results returned by the npx skills find command.
Ingestion points: Search results from the npx skills find [query] command in SKILL.md.
Boundary markers: None; external search results are mixed directly into the agent's context.
Capability inventory: Shell command execution (npx skills) and code installation capabilities are present in SKILL.md.
Sanitization: No sanitization or validation of the external search results or package names is performed before they are used to execute commands.

find-skills