The Agent Skills Directory

[REMOTE_CODE_EXECUTION] (HIGH): The skill provides instructions to run npx skills add <owner/repo@skill> -g -y, which downloads and executes code from arbitrary GitHub repositories.
Evidence: The -y flag explicitly skips confirmation prompts, enabling silent execution of remote code.
Evidence: Use of npx dynamically fetches and executes the skills package from the npm registry.
[COMMAND_EXECUTION] (HIGH): The skill grants the agent the ability to execute shell commands to manage system-wide software installations.
Evidence: Detailed instructions for npx skills find and npx skills add in SKILL.md encourage the agent to execute subprocesses based on external input.
[EXTERNAL_DOWNLOADS] (HIGH): The skill facilitates downloading packages from unverified external sources. While it mentions a trusted repository (vercel-labs/agent-skills), the command structure allows any GitHub user/repo to be targeted.
[PROMPT_INJECTION] (HIGH): The skill is highly vulnerable to Indirect Prompt Injection (Category 8) because it processes and acts upon untrusted data from an external registry.
Ingestion points: Search results from npx skills find [query] (sourced from skills.sh).
Boundary markers: Absent; search results are presented and processed as natural language instructions for the agent.
Capability inventory: Shell execution (npx skills add) and environment modification.
Sanitization: Absent; the agent is instructed to present findings and offer installation without verifying the content of the search results.

find-skills