The Agent Skills Directory

[PROMPT_INJECTION]: Potential for indirect prompt injection via external data processing.
Ingestion points: The skill frequently reads data from the environment via helm get manifest, kubectl logs, and kubectl get events as described in both SKILL.md and REFERENCE.md.
Boundary markers: No explicit boundary markers or instructions to ignore embedded instructions are used when processing external command outputs.
Capability inventory: The skill utilizes the Bash tool to perform administrative cluster operations using helm and kubectl binaries.
Sanitization: No sanitization or validation of external output is specified before the agent processes the information.
Remediation: Wrap external content in delimiters (e.g., XML tags) and provide the agent with explicit instructions to ignore any potential commands or guidance found within the processed output. Use structured data formats (like the -o json flag mentioned in the Agentic Optimizations section) to ensure data is parsed correctly rather than interpreted as natural language instructions.

helm-debugging