The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests and processes user-supplied YAML configuration files. There is a potential surface for indirect prompt injection if the files contain malicious instructions that influence the agent's summarization or triage logic.
Ingestion points: The skill reads a user-supplied YAML path (Step 1) and Kubernetes resources containing configRawYaml.
Boundary markers: The instructions lack explicit delimiters or 'ignore embedded instructions' warnings for the ingested YAML content.
Capability inventory: The agent has access to Bash(curl), Bash(jq), and Read tools, which are used to send the ingested data to external API endpoints.
Sanitization: No explicit sanitization or validation of the YAML content is mentioned before the agent summarizes it or interpolates it into terminal commands.
[COMMAND_EXECUTION]: The skill generates shell commands for curl and jq that incorporate user-provided variables such as CONFIG_PATH, QUERY, and TIMEZONE. This creates a surface for command argument injection if the agent fails to properly sanitize or quote these inputs when executing the shell commands.

vmanomaly-review