The Agent Skills Directory

[COMMAND_EXECUTION]: The skill facilitates the creation and execution of SHELL task nodes within data workflows. As documented in SKILL.md and references/dataops-flow-create.md, task definitions can include arbitrary bash scripts (e.g., ae-cli dataops_flow +save_task_definition ... --taskContentJson '{"script":"#!/bin/bash\necho hello"}'). This allows the agent to execute system commands on the environment where the orchestration engine runs.
[REMOTE_CODE_EXECUTION]: The skill enables the execution of dynamic SQL and DDL commands against remote data sources. Commands like ae-cli dataops_ide +execute_sql and ae-cli dataops_integration +execute_sql allow the agent to run potentially destructive operations on external databases if not properly constrained.
[CREDENTIALS_UNSAFE]: The skill is designed to manage database credentials for various integration components. While the provided documentation uses placeholders like xxx, the underlying functionality requires the agent to handle and store sensitive authentication data (host, port, username, password) when calling +add_datasource or +test_datasource_connect.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection as it ingests untrusted data from external sources.
Ingestion points: Query results fetched via +get_query_result (in references/dataops-query.md) and task logs retrieved via +get_task_instance_log (in references/dataops-flow-monitor.md).
Boundary markers: None identified. The instructions do not specify any delimiters or safety warnings to ignore instructions found within the processed data.
Capability inventory: The agent has extensive capabilities including shell execution, SQL execution (DML and DDL), and data synchronization management.
Sanitization: There is no evidence of sanitization, filtering, or validation of the data ingested from logs or database records before it is presented to the agent.

te-dataops