The Agent Skills Directory

[PRIVILEGE_ESCALATION]: The skill templates, specifically in template.md and examples/electron.md, include instructions for the agent to use sudo apt-get install to manage system-level packages, which requires administrative privileges and modifies the host environment.\n- [DYNAMIC_EXECUTION]: The Electron driver template provided in examples/electron.md implements a REPL with a dedicated eval command, enabling the execution of arbitrary JavaScript within the application's browser context using Playwright's page.evaluate() mechanism.\n- [COMMAND_EXECUTION]: The skill's operational instructions and examples make extensive use of shell commands for environment setup, process management (using tmux, kill, and pkill), and application execution within virtual framebuffers via xvfb-run.\n- [INDIRECT_PROMPT_INJECTION]: The skill is designed to process untrusted project data (such as READMEs, build manifests, and existing skill files) to generate new agent instructions. It lacks explicit boundary markers or sanitization for this ingested data, creating a potential surface for prompt injection from malicious content within the source project's documentation.

run-skill-generator