The Agent Skills Directory

[COMMAND_EXECUTION]: The skill includes preamble and telemetry blocks that execute bash commands automatically to read configuration files and collect system metadata like platform architecture.
[EXTERNAL_DOWNLOADS]: The skill uses curl to interact with external APIs for telemetry and protocol health checks.
[REMOTE_CODE_EXECUTION]: Instructions recommend installing community skills via npx, which involves downloading and executing code from external GitHub repositories.
[DATA_EXFILTRATION]: Local configuration and system environment details are collected and transmitted to a remote endpoint via curl. The skill also reads data from sensitive directories outside the standard workspace, specifically ~/.claude/ and ~/.superstack/.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it processes external JSON catalog files which could contain malicious instructions, and it lacks boundary markers or sanitization logic while maintaining access to shell execution tools.

navigate-skills