The Agent Skills Directory

[COMMAND_EXECUTION]: The skill's bootstrap workflow directs the agent to install the cloc utility using system package managers with sudo privileges (e.g., sudo apt install cloc, sudo yum install cloc). While standard for software installation, this grants the agent elevated permissions on the host system.
[PROMPT_INJECTION]: The skill contains an indirect prompt injection surface by instructing the agent to "Read the nearest AGENTS.md and confirm scope and constraints" and "respects the repo's AGENTS.md commands first." This allows an attacker who controls the repository content to potentially hijack the agent's behavior through instructions in the AGENTS.md file.
Ingestion points: AGENTS.md file in the repository being analyzed.
Boundary markers: None; the instructions explicitly tell the agent to prioritize content from this external file.
Capability inventory: Shell command execution through cloc and system package management tools.
Sanitization: No sanitization or validation of the commands in AGENTS.md is specified.
[EXTERNAL_DOWNLOADS]: The skill provides instructions for downloading and installing cloc from official and well-known sources, such as the AlDanial/cloc GitHub repository, official package registries (npm, Homebrew, Chocolatey, etc.), and Docker Hub (aldanial/cloc). These are legitimate distribution channels for the intended tool.

cloc