The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from infrastructure files and external web search results, which creates a potential surface for instructions hidden in data to influence agent behavior.
Ingestion points: The skill reads local *.tf and *.yml files during the discovery and dimension check steps, and fetches external information via WebSearch.
Boundary markers: There are no instructions in SKILL.md to use delimiters or 'ignore embedded instructions' prompts when the agent interpolates the content of discovered files or search results into its context.
Capability inventory: The skill has access to Bash, Read, and WebSearch tools, allowing for command execution and network operations that could be abused if malicious instructions in data are followed.
Sanitization: The skill lacks logic to validate, filter, or escape the content of external infrastructure files or search results before processing them.

devops-verifier