The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill enables the ingestion and processing of external Helm chart manifests, which represents a potential attack surface for indirect instructions.
Ingestion points: Content is ingested from local chart directories (Chart.yaml, values.yaml, and template files) by the scripts/validate-chart.sh script and Helm CLI commands.
Boundary markers: None; the skill does not define specific delimiters to isolate external manifest content from agent instructions.
Capability inventory: High-privilege tools are employed, including helm for rendering and dry-run installs, kubectl for resource inspection, and aws for repository synchronization.
Sanitization: While the skill promotes linting and schema validation for data structure, it lacks mechanisms to sanitize natural language instructions embedded within the processed manifests.
[EXTERNAL_DOWNLOADS]: The skill facilitates downloading chart dependencies from external sources.
Evidence: The documentation and Chart.yaml templates reference fetching subcharts (e.g., PostgreSQL, Redis) from the well-known Bitnami repository (https://charts.bitnami.com/bitnami) via the helm dependency update command.

helm-chart-scaffolding