The Agent Skills Directory

EXTERNAL_DOWNLOADS (LOW): The skill downloads Helm charts from the Bitnami repository using helm pull. While Bitnami is a reputable source, it is not included in the predefined trusted organizations list. This finding is downgraded to LOW as it is essential for the skill's primary function.
COMMAND_EXECUTION (LOW): The skill executes multiple shell commands (helm, git, docker compose) to validate and apply upgrades. It runs helm lint and helm template on downloaded third-party content, which processes external template logic. This is downgraded to LOW per the primary skill purpose rule.
PROMPT_INJECTION (LOW): The skill exhibits an indirect prompt injection surface (Category 8) by processing data from external GitHub repositories.
Ingestion points: Steps in chart-upgrade-process.md and image-upgrade-process.md involve reading Bitnami's CHANGELOGs and Dockerfiles on GitHub.
Boundary markers: Absent. The agent is instructed to read external content and extract specific values without clear delimiters or instructions to ignore embedded text.
Capability inventory: Subprocess execution (helm, docker), repository-wide file writing, and modification of CI/CD workflows in .github/workflows/.
Sanitization: The skill implements a version compatibility check to block major version jumps, but it does not sanitize or filter the content of the external documentation it reads.

upgrading-chart