The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides instructions for executing shell commands to audit, move, and delete files (e.g., rm -rf, git rm, cp). These are standard administrative actions consistent with the skill's stated purpose of repository cleanup.
[DATA_EXFILTRATION]: While the skill interacts with sensitive directory patterns such as .env and configuration folders, it contains no network operations (no curl, wget, or socket calls), posing no risk of exfiltration in its current form.
[INDIRECT_PROMPT_INJECTION]: The skill acts on data from the local repository (filenames, directory structures, and file content via grep and diff). This presents a standard surface for indirect prompt injection if a malicious repository contains specifically crafted filenames or content, though the commands used are robust utilities.
Ingestion points: find, ls, grep, diff, stat commands reading from the local filesystem.
Boundary markers: No specific delimiters are used for the shell command outputs.
Capability inventory: File deletion (rm, git rm), file copying (cp), and directory creation (mkdir).
Sanitization: No explicit sanitization of filenames before passing to shell commands is described.

hidden-folder-audit