The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses shell scripts and the sf CLI to orchestrate Salesforce metadata deployments and data operations. This is the intended primary purpose of the skill.
Evidence: scripts/configure-and-report.sh executes several sf commands such as sf data query, sf project deploy start, and sf api request rest.
Evidence: scripts/tests/_bootstrap.py uses subprocess.run to execute the shell script during testing.
[DYNAMIC_EXECUTION]: The skill dynamically generates Salesforce metadata XML files and configuration files in a temporary directory before deploying them to the target environment.
Evidence: scripts/configure-and-report.sh constructs a .skill-meta.xml file using a HEREDOC and interpolates a validated skill label.
[INDIRECT_PROMPT_INJECTION]: The skill processes external CSV data which is used in SOQL queries and metadata generation, representing an indirect injection surface.
Ingestion points: The agent_members_csv input in scripts/configure-and-report.sh accepts a comma-separated list of usernames or Salesforce IDs.
Boundary markers: No explicit prompt boundaries are used, but the input is strictly validated via regex before use.
Capability inventory: The skill can perform SOQL queries, Metadata API deployments, and REST DML (Insert) via the sf CLI.
Sanitization: Inputs are sanitized using a strict regex (allowing only 15/18-char alphanumeric IDs or standard email patterns). The skill label is escaped using the xml_escape function to prevent XML injection.

service-omni-skills-based-routing-configure