The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to source coding standards, conventions, and architectural patterns from external project files such as documentation, configuration files, and domain-specific markers. This creates a surface where malicious instructions embedded in those files could be interpolated into a high-level goal prompt and subsequently processed by execution tools.
Ingestion points: dna/*, docs/, .lev/validation-gates.yaml, execution.yaml, and project .lev/exec-profiles/.
Boundary markers: Absent. The template provided does not specify delimiters or instructions to ignore embedded commands within the sourced project data.
Capability inventory: The skill facilitates the use of the /exec tool, which is a bounded execution surface capable of running code and system commands.
Sanitization: The instructions do not define sanitization or validation logic for the content retrieved from external files.
[COMMAND_EXECUTION]: The skill directs the agent to execute shell commands to determine system state and tool capabilities. Specifically, it instructs running lev exec --help to verify CLI flags and lev exec "binding smoke" --dry-run to resolve environment bindings before dispatching tasks.

goal-exec