The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest untrusted data from external sources including runbooks, Docker Compose files, Kubernetes manifests, and Slack history.
Ingestion points: The scripts extract-from-runbook.sh, docker-compose-to-skill.sh, and k8s-to-skill.sh read input from external files.
Boundary markers: The generated skills include a disclaimer advising users to review and refine the output before use.
Capability inventory: The tool writes files to the local filesystem and generates templates containing ssh, kubectl, and docker-compose commands.
Sanitization: There is no evidence of security-focused sanitization or instruction filtering of the input data before it is incorporated into new skill files.
[COMMAND_EXECUTION]: The provided utility scripts execute system commands such as mkdir, find, and cat, and utilize external utilities like yq to process and transform YAML configuration files.
[DYNAMIC_EXECUTION]: The skill programmatically generates new markdown-based skill files (SKILL.md) by extracting, concatenating, and transforming content from user-provided documentation and infrastructure manifests.

infrastructure-skill-builder