The Agent Skills Directory

[PRIVILEGE_ESCALATION]: The skill contains instructions for users to execute high-privilege commands to manage system security settings.
Evidence: SKILL.md provides the command sudo security add-trusted-cert -d -r trustRoot -k /Library/Keychains/System.keychain openchoreo-ca.crt for macOS users to trust self-signed certificates.
[INDIRECT_PROMPT_INJECTION]: The skill has an attack surface where instructions fetched from external or untrusted sources could influence the agent's behavior.
Ingestion points: Remote documentation pages fetched by scripts/fetch-page.sh from openchoreo.dev and workload.yaml descriptors provided in application source repositories.
Boundary markers: No explicit delimiters or instructions to ignore embedded commands were found when processing fetched content.
Capability inventory: The skill uses occ, kubectl, git, and gh to perform cluster and repository operations.
Sanitization: The skill lacks explicit sanitization for the content of documentation pages or workload descriptors before they are parsed or acted upon.
[EXTERNAL_DOWNLOADS]: The skill uses utility scripts to fetch data from the vendor's infrastructure and GitHub repositories.
Evidence: scripts/fetch-page.sh and scripts/list-versions.sh download content from openchoreo.dev and raw.githubusercontent.com/openchoreo/* to retrieve documentation and version info.
[COMMAND_EXECUTION]: The skill's primary operation involves executing local CLI tools to manage infrastructure as code.
Evidence: Frequent use of occ (OpenChoreo CLI), kubectl, flux, git, and gh across all recipe files and the main skill instructions.

openchoreo-developer-gitops