The Agent Skills Directory

[DYNAMIC_EXECUTION]: The skill includes a bootstrap script (scripts/invoke.py) that dynamically searches for a local manager tool (kntnt) across a wide variety of standard AI agent configuration directories in the user's home folder. Once found, it executes the manager's engine using uv run. This is a mechanism designed to provide platform-agnostic integration for the skill framework.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process local project files, such as deployment documentation and infrastructure configurations. This creates a surface area for indirect prompt injection if those project files contain malicious instructions.
Ingestion points: The skill reads deployment documentation, existing infrastructure files, and configuration/release files as specified in SKILL.md and references/deployment.md.
Boundary markers: None identified.
Capability inventory: The skill possesses the capability to execute commands via subprocess.run (to invoke the manager) and is authorized by its instructions to use hcloud, ssh, and cloud-init for remote operations.
Sanitization: The skill instructions explicitly direct the agent to filter debug logs, structured data, and application logs for secrets before returning output to the context.
[SAFE]: The skill documentation demonstrates high security awareness by instructing the agent to keep credentials out of chat and committed files, verify SSH host-key fingerprints through trusted channels, and obtain specific authorization before performing destructive or billed actions.

hetzner