The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill uses curl -sL "..." | bash to download and execute a remote script (audit-scan.sh) from a GitHub repository. This pattern allows for arbitrary code execution on the user's machine if the remote source is compromised.
[COMMAND_EXECUTION]: The skill executes multiple shell commands to inspect the local environment, including curl, cat, ls, and jq.
[CREDENTIALS_UNSAFE]: The skill reads sensitive configuration files such as ~/.claude/settings.json and ~/.claude.json. These files often contain security-sensitive information, including tool permissions, environment variables, and configurations for MCP servers that may store API keys or tokens.
[EXTERNAL_DOWNLOADS]: The skill fetches a large remote knowledge base (up to 3000 lines of ultimate-guide.md) via curl, which is then processed as instructions for the agent.
[INDIRECT_PROMPT_INJECTION]: The skill has a significant attack surface for indirect prompt injection:
Ingestion points: Fetches ultimate-guide.md and audit-scan.sh from a remote GitHub repository.
Boundary markers: None. The content of the remote markdown file is treated as a direct knowledge reference without delimiters.
Capability inventory: The skill has the ability to execute shell commands, read local configuration files, and pipe remote content to bash.
Sanitization: No sanitization or validation is performed on the remote content before execution or processing.

diagnose