The Agent Skills Directory

[SAFE]: The skill implements read-only functionality using standard tools like rg, sed, and nl. These are used for searching and reading files within the project scope, which is the intended purpose of the skill.
[SAFE]: Explicit guardrails are provided in the instructions to prevent the agent from creating files, modifying the codebase, running installers (npm install), or using write-oriented commands (mkdir, git commit). This minimizes the risk of accidental or malicious state changes.
[SAFE]: The workflow focuses on reconnaissance and summarizing findings rather than executing code, effectively reducing the risk of indirect prompt injection from codebase content.

lightweight-explorer