The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute a local PowerShell script located at scripts/oracle-gate.ps1 and a CLI tool named oracle to perform its primary functions.
[DATA_EXFILTRATION]: By design, this skill is intended to send local file content and task context to an external expert service (Oracle). The instructions explicitly warn against sending secrets, credentials, customer data, or production logs, providing a 'dry-run' step to review files before they are transmitted.
[INDIRECT_PROMPT_INJECTION]: The agent is instructed to read and process output generated by an external expert model. This creates a surface where the external model's response could influence the agent's subsequent actions.
Ingestion points: The agent reads the final report or output from the Oracle consultation (e.g., Markdown files in .oracle-runs/).
Boundary markers: The skill specifies a structured 'Result Format' for the agent to use when summarizing findings (Accepted/Rejected/Needs verification), which acts as a logical delimiter for the external data.
Capability inventory: The skill has the capability to execute shell commands via PowerShell and the oracle CLI tool.
Sanitization: There are no explicit sanitization routines for the external output, though the agent is instructed to 'Verify locally' before treating findings as final.

oracle-consult