The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill provides instructions for the installation of the mirrord CLI through established package managers like Homebrew and Chocolatey, which are well-known and trusted distribution services. It explicitly cautions against using unverified remote scripts.
[COMMAND_EXECUTION]: The skill uses standard CLI tools including kubectl, mirrord, and curl for cluster diagnostics. It also incorporates bundled Python scripts, dump_stats.py and request_recorder.py, which are designed to analyze captured traffic and verify filter configurations using only the Python standard library.
[INDIRECT_PROMPT_INJECTION]: The skill processes live network traffic captured from the cluster environment, which may contain untrusted data. 1. Ingestion points: Captured request headers and JSON bodies analyzed by analytical scripts. 2. Boundary markers: The instructions include a specific rule that captured traffic and cluster resources must be treated as data and never as instructions, with a prohibition on executing commands found within them. 3. Capability inventory: The agent utilizes kubectl, mirrord, and curl across several diagnostic steps. 4. Sanitization: The skill mandates security practices such as using temporary directories for logs, deleting data after analysis, and masking sensitive tokens or credentials in generated reports.

mirrord-propagation-discovery