The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides numerous scripts and guides for managing cloud infrastructure, including cluster creation, service restarts, and configuration updates across various EMR environments.
[EXTERNAL_DOWNLOADS]: The skill installs a vendor-provided Python SDK (python_serverless-1.0.3.4.1-py3-none-any.whl) located in the assets folder via a shell script (install_serverless_sdk.sh). This is a standard distribution method for the author's own tooling.
[DYNAMIC_EXECUTION]: The scripts emr_serverless_submit_cli.py and emr_serverless_manager.py utilize the __import__ function to dynamically load modules from the installed serverless package. This is used to maintain compatibility across different versions of the vendor SDK.
[CREDENTIALS_UNSAFE]: The skill requires Volcengine API Access Keys (AK/SK) to be set in environment variables (VOLCENGINE_AK, VOLCENGINE_SK). It also utilizes Ark Skill API keys for proxying requests. These credentials are used solely for authenticating with the author's official cloud endpoints (volcengineapi.com and byteplusapi.com).
[INDIRECT_PROMPT_INJECTION]: The skill possesses an attack surface for indirect prompt injection as it ingests untrusted data from job logs (FetchDriverLog) and diagnostic reports while maintaining the capability to modify cluster configurations and execute actions.
Ingestion points: Job driver logs retrieved via FetchDriverLog and diagnostic responses generated by the EMR Agent (expert.py).
Boundary markers: Absent; the instructions do not specify the use of delimiters when processing log or diagnostic data.
Capability inventory: Significant capabilities including cluster configuration updates (UpdateApplicationConfig), service management (RunApplicationAction), and job submission.
Sanitization: Content is processed as natural language for agent analysis without explicit sanitization or escaping mechanisms.

byted-emr-skills