The Agent Skills Directory

[PROMPT_INJECTION]: The skill exhibits an Indirect Prompt Injection attack surface because it processes data from external webhook providers (Front, GitHub, Vercel, Todoist, Mux) and passes it to an LLM inference utility. * Ingestion points: Webhook handlers in src/webhooks/providers/ and Inngest event payloads. * Boundary markers: No explicit markers or 'ignore instructions' warnings are specified for the infer prompt interpolation. * Capability inventory: Subprocess execution via Bun.spawn (CLI tools pi, joelclaw), Kubernetes management via kubectl, and network operations. * Sanitization: No validation or sanitization of incoming webhook payloads is described.
[COMMAND_EXECUTION]: The skill relies on executing various local commands and management scripts for its primary purpose. * Executes local deployment scripts located at ~/Code/joelhooks/joelclaw/k8s/publish-system-bus-worker.sh. * Spawns subprocesses for the pi CLI to perform inference and the joelclaw CLI for infrastructure management. * Uses kubectl for container orchestration tasks.

system-bus