The Agent Skills Directory

[SAFE]: The skill implements management of Volcano Engine cloud resources using official Bytedance/Volcano Engine APIs and the volcengine-python-sdk. All external network interactions target expected cloud provider endpoints or user-specified Elasticsearch/OpenSearch clusters.- [SAFE]: Secret management is handled correctly via environment variables (e.g., VOLCENGINE_ACCESS_KEY, ESCLOUD_API_KEY). The documentation explicitly instructs users to set these in their environment, avoiding hardcoded credentials.- [SAFE]: Destructive operations, such as cluster deprovisioning in scripts/control.py and index cleanup in scripts/data.py, are protected by mandatory --confirm CLI flags and explicit agent guardrails in SKILL.md to prevent accidental or unauthorized resource destruction.- [INDIRECT_PROMPT_INJECTION]: The skill possesses an attack surface for indirect prompt injection as it retrieves and processes data from external search clusters (e.g., document contents, index mappings).
Ingestion points: Data entering the agent via scripts/data.py (e.g., index_get, smoke_test, cat_nodes).
Boundary markers: The SKILL.md and DATA_PLANE.md files provide explicit routing and decision rules to the agent to maintain focus on specific tasks.
Capability inventory: The skill can execute cloud lifecycle mutations (provision, delete, scale) and data-plane operations (index, search, delete) via subprocess calls to the bundled Python scripts.
Sanitization: Standard library-level sanitization is provided by the opensearch-py client and the Volcano Engine SDK, though semantic-level safety relies on the agent's core instructions and the skill's documented guardrails.

byted-escloud