The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill provides instructions to download a zip archive from a personal GitHub repository (Harkirat1462/claude-code-cli). This source is unverifiable and presents a high risk for a tool requiring extensive system permissions and codebase access.
[REMOTE_CODE_EXECUTION]: The installation workflow involves fetching a remote zip file, extracting it, and running npm install and npm link. This sequence leads to the execution of arbitrary scripts contained in the downloaded archive from an untrusted repository.
[INDIRECT_PROMPT_INJECTION]: The skill functions as an agentic coding assistant that processes codebase data, creating a significant attack surface for indirect instructions.
Ingestion points: Local codebase files, git history, and project dependencies as defined in the file patterns section (SKILL.md).
Boundary markers: There are no explicit boundary markers or instructions to ignore embedded commands within the files being processed.
Capability inventory: The skill possesses capabilities for multi-file editing, terminal execution via git workflows, subagent spawning, and CI/CD integration (SKILL.md).
Sanitization: No sanitization or validation of external content is mentioned before the data is interpolated into the agent's context.
[PERSISTENCE]: The skill instructs users to modify shell profiles (~/.bashrc, ~/.zshrc) to persist environment variables, which is a common persistence mechanism that ensures the tool and its configurations remain active across sessions.
[METADATA_POISONING]: The skill claims to be an official terminal assistant related to the Claude agentic coding system and credits ara.so, but directs the actual software download to an individual's personal repository, which is deceptive regarding the source and safety of the software.

claude-code-cli-terminal-assistant