The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill uses bunx to download and execute the cc-safety-net package from the npm registry for status reporting, diagnostics, and command analysis. This represents an unverified remote code execution pattern.
[EXTERNAL_DOWNLOADS]: Installation instructions require downloading plugins and extensions from unknown external repositories, including https://github.com/kenryu42/gemini-safety-net and a custom marketplace kenryu42/cc-marketplace. These sources are not affiliated with the skill author or a trusted vendor.
[COMMAND_EXECUTION]: The skill's core functionality relies on the execution of external CLI tools that intercept and modify shell command behavior across multiple platforms. This puts the unverified code in a high-privilege position between the user and the system.
[INDIRECT_PROMPT_INJECTION]:
Ingestion points: The skill ingests and analyzes all shell commands (untrusted data) through a PreToolUse hook (e.g., in SKILL.md).
Boundary markers: There are no mentioned delimiters or safety instructions to prevent the analyzer from being manipulated by the commands it processes.
Capability inventory: The skill has the capability to block any command execution and uses bunx to execute arbitrary packages.
Sanitization: While the skill claims to redact sensitive data, the logic used to analyze and filter commands is contained in an external, unverified package.

claude-code-safety-net