The Agent Skills Directory

[PROMPT_INJECTION] (SAFE): The instructions in SKILL.md define a benign workflow for keyword-based resource selection and retrieval. There are no attempts to override system prompts, bypass safety filters, or extract system instructions.- [REMOTE_CODE_EXECUTION] (SAFE): Resource '10-distribution.md' describes common distribution patterns like 'curl | bash'. These are clearly labeled as developer guidelines and examples for third-party tool installation, not code to be executed by the agent or the skill.- [COMMAND_EXECUTION] (SAFE): The guideline files contain various code snippets (Python, Bash, etc.) demonstrating how to use subprocesses for pagers or editors. These are static examples and do not constitute a capability of the skill to execute commands on the host.- [DATA_EXFILTRATION] (SAFE): There are no active network calls or file system access operations defined in the skill's logic. Mentions of sensitive files or environment variables (like MYAPP_PASSWORD) are used as negative examples or generic placeholders in documentation.- [NO_CODE] (SAFE): The skill consists entirely of Markdown files. It does not include any scripts, executables, or configuration files that would be run in the agent's execution environment.

cli-guidelines