The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill instructs the agent to download and execute a shell script from an external URL by piping it directly into bash.
Evidence: curl -fsSL https://api.gtm-tools.sh/extension/install.sh | bash in SKILL.md.
[EXTERNAL_DOWNLOADS]: The skill downloads executable content from a non-whitelisted domain to install an extension.
Evidence: https://api.gtm-tools.sh/extension/install.sh.
[COMMAND_EXECUTION]: The skill extensively uses shell-based tool invocations and CLI commands for setup and operation.
Evidence: Multiple curl commands for authentication and tool execution, plus gtm-tools admin login for CLI authentication.
[DATA_EXFILTRATION]: The skill manages and transmits sensitive data, including API keys and user interaction data, to an external API service.
Evidence: Instructions for handling GTM_TOOLS_API_KEY and calls to api.gtm-tools.sh for identity registration and data retrieval.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it ingests and processes untrusted data from external sources (LinkedIn and Reddit) that could contain malicious instructions.
Ingestion points: Tools such as get_linkedin_post, get_reddit_post, and list_user_posts in references/tools.md retrieve external content.
Boundary markers: No delimiters or "ignore instructions" warnings are present in the provided documentation.
Capability inventory: The skill has significant capabilities including shell execution (curl, bash) and file system access (~/.gtm-tools/config.json).
Sanitization: No sanitization or validation steps are described for processing external post content.

gtm-tools