The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill provides a standard installation method for Linux that involves fetching and executing a shell script from the official Helm project's GitHub repository using curl | bash in references/install-and-setup.md.
[EXTERNAL_DOWNLOADS]: Recommends the installation of several community-maintained plugins and tools from external GitHub repositories, including the asdf-helm plugin, helm-diff, and helm-secrets.
[COMMAND_EXECUTION]: The skill involves extensive use of helm and kubectl commands to manage and inspect Kubernetes cluster resources, including release installation, upgrades, and history tracking.
[COMMAND_EXECUTION]: Includes commands requiring administrative privileges via sudo for system-level tasks such as adding repository signing keys, installing packages through the system package manager, and setting up shell completion scripts in protected system directories.
[PROMPT_INJECTION]: The skill processes Helm charts and templates (e.g., NOTES.txt and manifest templates) that ingest external configuration values. This creates a surface for indirect prompt injection where malicious instructions embedded in a chart's data could potentially influence the agent's behavior.
Ingestion points: Chart.yaml, values.yaml, and template files located in the templates/ directory of a chart as described in references/chart-authoring.md.
Boundary markers: The skill does not explicitly define delimiters to isolate processed template content from the rest of the agent's context.
Capability inventory: The skill has broad capabilities for executing shell commands via helm and kubectl across multiple reference files.
Sanitization: No specific sanitization or validation of external chart content is mandated; the skill relies on Helm's native Go template rendering engine.

helm