The Agent Skills Directory

[SAFE]: The skill acts as a documentation bridge for Claude Code, correctly identifying that training data may be stale and prioritizing live configuration and official documentation.
[EXTERNAL_DOWNLOADS]: The skill references and fetches content from official and trusted domains: code.claude.com, claude.com, claude.ai, and platform.claude.com. These are legitimate vendor resources used for providing up-to-date documentation.
[PROMPT_INJECTION]: The skill uses instructional language to direct the agent to prioritize specific information sources (live documentation over training data). These instructions are for maintaining information accuracy and do not represent a security bypass or malicious attempt to override safety filters.
[CREDENTIALS_UNSAFE]: No hardcoded credentials or unsafe secrets management were found. The skill correctly mentions standard configuration paths like ~/.claude/settings.json and ~/.claude/keybindings.json for user-directed configuration.
[DATA_EXFILTRATION]: No data exfiltration patterns were detected. Network access is restricted to fetching documentation from verified vendor domains.
[COMMAND_EXECUTION]: The skill mentions commands like claude --help and /install-slack-app as suggestions for the user to run in their own terminal, but it does not attempt to execute arbitrary or dangerous shell commands autonomously.

claude-code-docs