The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructs the agent to download an installation script from the vendor's official website at https://ryvn.ai/install.sh to set up the environment.
[REMOTE_CODE_EXECUTION]: The skill uses a curl | bash pattern to execute the vendor's installation script for the ryvn CLI.
[COMMAND_EXECUTION]: The skill utilizes the system shell to run ryvn CLI commands for infrastructure management, resource provisioning, and log retrieval.
[PROMPT_INJECTION]: Instructions in the Feedback section of SKILL.md encourage the agent to automatically submit feedback to the vendor without seeking user permission when encountering bugs or workflow issues, which reduces user oversight for that specific action.
[INDIRECT_PROMPT_INJECTION]: The skill provides an attack surface for indirect prompt injection by ingesting and processing external, potentially untrusted data from application logs (ryvn logs) and Kubernetes manifests (ryvn get manifest).
Ingestion points: Reading logs and Kubernetes manifests in operate.md.
Boundary markers: None explicitly defined in the instructions for log processing.
Capability inventory: Full access to Bash and the ryvn CLI to manipulate infrastructure.
Sanitization: Relies on the underlying CLI and LLM guardrails.

use-ryvn