The Agent Skills Directory

[COMMAND_EXECUTION]: The health check script (scripts/health.sh) executes numerous system utilities such as kubectl, talosctl, git, bun, df, and du to collect diagnostic data.
[COMMAND_EXECUTION]: The skill's documentation provides remediation commands for the agent or user that involve destructive actions, such as rm -rf and find ... -delete, specifically targeting temporary files and stale test artifacts within the project directory.
[REMOTE_CODE_EXECUTION]: The diagnostic process involves running bun test and bunx tsc --noEmit, which executes code present in the local repository packages.
[CREDENTIALS_UNSAFE]: The script accesses local configuration in ~/.talos/config and uses a local secrets utility to retrieve a keyring password for the gogcli tool to verify Google Workspace authentication status.
[EXTERNAL_DOWNLOADS]: The skill performs a git fetch operation to check for synchronization status between the local repository and its remote origin.
[PROMPT_INJECTION]: The skill is identified as having an attack surface for indirect prompt injection (Category 8) because it processes and displays output from external tools and local test suites directly into the agent's context.
Ingestion points: Diagnostic output from joelclaw, kubectl, git, and bun test in scripts/health.sh is ingested by the agent.
Boundary markers: No explicit delimiters or warnings (e.g., "ignore embedded instructions") are present in the script's output formatting.
Capability inventory: The skill environment possesses capabilities including file deletion, infrastructure management via kubectl, and package installation via bun.
Sanitization: No sanitization or validation of tool output is performed before presentation to the agent.

joelclaw-system-check