The Agent Skills Directory

[DYNAMIC_EXECUTION]: The skill uses vm.runInContext in references/profiles/env-patch/scripts/env-diagnose.js to execute user-supplied JavaScript code for diagnostic purposes. Additionally, exec() is used in tests/test_scaffold_reverse_project.py to load and verify generated collector logic during test runs.
[COMMAND_EXECUTION]: Utility and validation scripts, such as scripts/check_reverse_env.py and scripts/validate_skill.py, use subprocess.run() to execute shell commands for checking system tool versions and running internal self-tests.
[INDIRECT_PROMPT_INJECTION]: The skill's primary workflow involves ingesting and processing untrusted third-party data, including HAR transcripts and JavaScript source code. This ingestion surface, combined with the skill's built-in local execution capabilities, represents a risk if malicious artifacts are analyzed without using an external security boundary or sandbox as recommended in the documentation.
[REMOTE_CODE_EXECUTION]: The skill is designed to analyze and execute (potentially hostile) JavaScript code provided by the user within a simulated browser environment. While it includes warnings about the safety of the Node.js vm module and suggests sandboxing, the inherent capability to run untrusted code locally represents an execution risk.

spider-king